Control of Thymic T Cell Maturation, Deletion and Egress by the RNA-Binding Protein HuR¹

Thymic T cell development involves discrete stages of differentiation, activation, death, and migration aiming to provide a T cell repertoire against invasion while maintaining tolerance to self. Double-negative (DN)³ progenitors (CD4⁻CD8⁻) entering the thymic cortex from the bone marrow proceed through four differentiation stages to yield cells competent for antigenic stimulation. In the case of conventional T cells, this is achieved by the somatic recombination of TCRβ genes, the elimination of aberrant rearrangements, and the association of the TCRβ-chains with the invariant TCRα-chain. This pre-TCR complex signals DN cells to expand and become CD4⁺CD8⁺ double-positive (DP) cells (1). DP cells rearrange their TCRα locus and progressively express surface TCRα/β for interaction with MHC-presented Ags on the thymic epithelium. The qualitative and quantitative features of these interactions guide the selection of nonself reacting CD4⁺ or CD8⁺ single-positive (SP) thymocytes (2). Nominal or strong autoreactive interactions cause elimination by neglect or negative selection, respectively; moderate interactions lead to positive selection. In parallel, an array of cytokine/chemokine signals define the thresholds of these interactions and facilitate thymocyte movements from the cortex to the medulla and peripheral blood (3, 4). These staged events require the precise orchestration of gene expression programs governed by genetic, epigenetic, and transcriptional mechanisms.

Posttranscriptional mechanisms of mRNA use may also affect thymocyte development because they contribute to as much as 50% of T cell-specific gene expression changes during activation (5). However, the trans-acting factors regulating these changes have not been well studied. Such factors include the RNA-binding proteins (RBPs) that determine mRNA maturation, localization, stability, and translation (6). This type of control is particularly stringent in mRNAs that possess adenylate and/or uridylate-rich elements and frequently encode cytokines, signaling molecules, and oncoproteins (7). Within the selective set of RBPs recognizing adenylate and/or uridylate-rich elements, HuR (or HuA) emerged as a pleiotropic modulator of mRNA use. HuR contains RNA recognition motifs with an affinity for a U-rich motif and it is the only ubiquitous member of the otherwise neuronal Elavl/Hu family (8). HuR shuttles between the nucleus and the cytoplasm via interactions with nuclear export/import adaptor proteins (9). In the cytoplasm, HuR can affect mRNA stability and/or translation (8, 10, 11, 12) via complex interplays with other RBPs, such as hnRNPD/AUF1, TTP, BRF1 and KSRP, as well as microRNAs (13, 14). Recent data suggest that HuR’s functions can be controlled by posttranslational modifications like phosphorylation, methylation, and cleavage (15, 16, 17, 18). Furthermore, HuR:RNA immunoprecipitations demonstrated that, in vivo, HuR associates with mRNAs that are relevant to a certain cellular response (6, 14), suggesting that its functions may vary in a tissue and signal-specific manner. Such an example has been provided for macrophages where HuR’s overexpression suppressed the translation of specific inflammatory mRNAs and attenuated acute inflammatory reactions (10). HuR’s involvement in adaptive immunity has been inferred by its interactions with mRNAs encoding CD3ζ, TNF, CD95/Fas, APRIL, CD40L, GM-CSF, IL-3, IL-4, and IL-13 (19, 20, 21, 22, 23, 24, 25), but these interactions were not informative on HuR’s control over multiparametric immune responses. The obligatory deletion of HuR in the mouse blocked embryonic development, thus prohibiting the assessment of its involvement in immune processes (26). The analysis of these embryos highlighted HuR’s control over developmental morphogenesis, but it could not reveal its effect on fundamental cellular responses. The staged progression of thymocyte development is an ideal process for assessing both HuR’s cellular functions and its role in adaptive immunity. In this study, we restricted HuR’s deletion in mouse thymocytes and revealed its control over distinct signal-dependant cellular responses.

The generation of the floxed Elavl1 allele (Elavl1^fl) was described in (26). Elavl1^fl/fl mice were backcrossed to C57BL/6J background for at least five generations. Additional transgenic mice were provided by J. D. Marth (LckCre; University of California, San Diego, CA) (27), D. Kioussis (F5-TCR; National Institute for Medical Research, London, United Kingdom) (28, 29), B. Malissen (TgHYTCR; Centre d’Immunologie de Marseille-Luminy, Université de la Méditerrannée, Marseille, France) (30), and G. Kollias (RAG1^−/− Biomedical Sciences Research Centre (BSRC) “Alexander Fleming”, Vari, Greece). All transgenic mice were crossed to C57BL/6 for more than nine generations. Congenic C57BL/6–CD45.1 mice were provided by A. Potocnik (National Institute for Medical Research, London, United Kingdom). All mice were bred and maintained in the animal facilities of the BSRC “Alexander Fleming” under specific pathogen-free conditions. Experiments on live animals were approved by the Hellenic Ministry of Rural Development (Directorate of Veterinary Services) and by BSRC Alexander Fleming’s Animal Research and Ethics Committee for compliance to Federation of European Laboratory Animal Science Associations’ regulations.

Cells from thymus, peripheral blood, spleen, and mesenteric lymph nodes were isolated using standard procedures. For stimulation, cells were cultured in complete RPMI 1640 with 2.5% FBS. For TCR-induced activation of kinases, thymocytes were first incubated on ice with anti-CD3 or anti-CD3/CD28 (0.5 μg/ml each) and then were cross-linked with goat anti-hamster IgG (40 μg/ml) at 37°C for 5 min. For HuR detection during apoptosis, thymocytes were stimulated with anti-Fas (100 ng/ml) in the presence of Cycloheximide (0.1 μM; Sigma-Aldrich) or Dexamethasone (100 nM; Sigma-Aldrich). For the chemokine-induced translocation of HuR, thymocytes were stimulated with CCL21 or SDF-1 (10 ng/ml; PeproTech) for 3 h.

For surface Ags, staining was performed via standard procedures. For intracellular Ags, cells were processed with Cytofix and Cytoperm (BD Biosciences) according to the manufacturer’s protocol. The flow cytometric detection of HuR was achieved by a biotin-conjugated anti-HuR Ab (3A2; sc5261) generated by Santa Cruz Biotechnology upon request, whereas the mIgG1 isotype control (IC002B) was purchased from R&D Systems. The remaining Abs for flow cytometry were purchased from BD Biosciences, with the exception of EDG-1 (Santa Cruz Biotechnology); CD4, T3.70 (HY-TCR), and CXCR4 (eBioscience); CCR7, CD45.1, and CD45.2 (BioLegend); Kruppel-like transcription factor 2 (Chemicon International); and biotinylated DR3 (R&D Systems). Stained cells were analyzed with FACS Canto II or sorted with using FACSVantage SE flow cytometer (BD Biosciences). Data were acquired with CellQuest or FACSDiva software (BD Biosciences) and analyzed using FlowJo (Tree Star). DN thymocytes were sorted as CD4⁻CD8⁻CD11b⁻CD45R/B220⁻CD49β⁻TER119⁻ after enrichment by negative selection using the IMag mouse CD4 lymphocyte enrichment set (BD Biosciences) followed by FACS sorting.

For in vivo cell cycle analysis, 8–10-wk-old mice were injected i.p. with 100 μl of 10 mg/ml BrdU (Sigma-Aldrich) and killed after 60 min. For BrdU staining, 5–10 × 10⁶ thymocytes were first stained for surface markers, then treated with Cytofix (BD Biosciences) for 1 h on ice and finally incubated with FastImmune anti-BrdU with DNase (BD Biosciences) for 1 h at room temperature. After DNA staining with 7-amino-actinomycin (5 μg/ml; Sigma-Aldrich), cells were analyzed by flow cytometry. For ex vivo proliferation assays, thymocytes were cultured in triplicates in microtiter plates (3 × 10⁴/well) and stimulated with plate-bound anti-CD3 (10 μg/ml), anti-CD3 plus anti-CD28 (10 μg/ml; BioLegend), or soluble PMA (10 ng/ml; Sigma-Aldrich) and ionomycin (500 ng/ml; Calbiochem). At 72 h, wells were pulsed with 1μCi [³H]thymidine and 6 h later they were harvested onto glass membranes for scintillation counting.

Bone marrow cells were isolated from femurs/tibia of 6–8-wk-old male CD45.1 (competitor), CD45.2 LckCre⁺ Elavl1^fl/+ (control), and CD45.2 LckCre⁺ Elavl1^fl/fl (test) mice. The representation of bone marrow progenitors was similar among genotypes as assessed by flow cytometry. Following purification, 4 × 10⁵ cells containing a 1/1 mixture from each genotype were injected into the tail-vain of irradiated (800 Gy), age-matched male CD45.2 RAG1-deficient mice. Reconstitution was monitored by flow cytometric analysis of peripheral blood. Mice were killed past the age of 8 wk for the estimation of the CD45.1/CD45.2 percentages in the thymus and peripheral lymphoid organs via flow cytometry. The corresponding CD45.2 percentages were normalized to CD45.1 and expressed as ratios of CD45.2 test:CD45.2 control.

Intrathymic FITC (10 μg/thymic lobe; Sigma-Aldrich) injections were performed as described (31). Chemotactic assays were performed in 5 μm Transwell filters (Corning), for 3 h to S1P (Avanti Polar Lipids), CCL21, CXCL12/SDF-1 (PeproTech), or medium in the bottom chamber and enumerated by flow cytometry, as described (32). Assays were performed in duplicate for each concentration, and were repeated using cells from a minimum of two different animals per genotype. Values were normalized to passively migrating DP cells.

For TCR-induced thymocyte apoptosis, 6–7 wk-old mice were injected i.p with either 25 μg anti-CD3 (145 2C11) or saline at 0 and 24 h. Twenty-four hours after the last injection, thymocytes were enumerated by flow cytometry. For the apoptotic response of F5 TCR-transgenic thymocytes, 6-wk-old mice were injected i.p. daily for 4 days with either saline or 20–50 nM NP68 peptide in saline. The NP68 peptide was the 9-mer Ala-Ser-Asn-Glu-Asn-Met-Asp-Ala-Met (366–374) from the nucleoprotein of influenza virus A/NT/60/68. On day 4, thymi, spleens, and mesenteric lymph nodes were removed for flow cytometry. For ex vivo apoptotic assays, 5 × 10⁶/ml thymocytes from 6–7-wk-old mice were cultured in plates coated with 1 μg/ml CD3 mAb and either 2 or 20 μg/ml CD28 mAb. Cells were harvested 24 h later and stained with anti-CD4, anti-CD8, and Annexin V for flow cytometry. For Fas-induced apoptosis, thymocytes were stimulated with various doses of anti-Fas Ab (Jo-2, 0–1000 ng/ml) plus 0.1 μM cycloheximide in RPMI 1640. Apoptosis in DP thymocytes was assessed 18 h later by Annexin V staining and flow cytometry.

Whole-cell lysates were prepared in RIPA buffer, whereas nuclear and cytoplasmic extracts were prepared with the NE-PER reagent (Pierce). Equimolar amounts of protein were analyzed on SDS-containing polyacrylamide gels and blotted onto nitrocellulose membranes (Schleicher & Schuell Microscience). Probing Abs included: HuR (3A2), procaspase-3 (l-18), actin (C-11), p27 (F-8), p21 (F-5), U1snRNP (C-18), Lck (3A5), ZAP70 (1E7.2), and PKCθ (C18) (all obtained from Santa Cruz Biotechnology); cleaved caspase-3 (Asp175), caspase-8, p-Src (Tyr 416, recognizing Lck phosphorylated at Tyr 394), pZAP70 (Tyr 319), pPKCθ (Thr538) and p53 (1C12) (all obtained from Cell Signaling Technology). Primary Abs were detected with HRP-conjugated secondary Abs (Southern Biotechnologies) and the ECL+ substrate (Amersham).

Total RNA was extracted from thymocytes using the RiboPure Kit (Ambion) according to the manufacturer’s instructions. RNA-immunoprecipitation (R-IP) assays were performed as previously described (10) using agarose conjugated anti-HuR (3A2; sc5261) or control IgG (sc2343) Abs generated by Santa Cruz Biotechnologies upon request. For RT-PCR analyses, 5–10 μg of cellular RNA were used for cDNA synthesis with Moloney murine leukemia virus reverse transcriptase (Promega). cDNAs were used for quantitative RT-PCR performed using Platinum SYBR Green qPCR SuperMix UDG (Invitrogen) on RotorGene 6000 machine (Corbett Research). Expression was normalized to B2 microglobulin and L32 mRNAs. Relative expression was calculated as the fold difference to control values that were assigned an arbitrary expression value of 1, using BioRad RelQuant or REST 2005 (Corbett Research). Control values were assigned an arbitrary expression value of 1. In all cases data were derived from three mice, groups, or genotypes analyzed in triplicate and summing up to nine independent reactions for each measurement. Primer sequences, microarray analyses, and meta-analyses are described in Extended Methods.⁴ The microarray data was submitted to the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo) under the record number GSE9174.

Unless otherwise indicated, Student’s t test was used for statistical analysis. Results with a p value <0.05 were considered significant.

To link HuR’s functions to thymocyte responses, we first assessed its abundance in thymocyte subsets. The highest HuR content was detected in DN thymocytes, particularly in DN3 (CD25⁺CD44⁻) cells (Fig. 1, A and B), relating to TCRβ selection and pre-TCR signaling. HuR levels dropped in DP cells, but rose again in SP subsets, correlating to TCR signals driving thymocyte selection. (Fig. 1, A and B). TCR agonists (anti-CD3/CD28), mitogens (PMA/ionomycin), and apoptotic stimuli (anti-Fas) increased the levels of HuR in SP subsets, which contain high surface TCR (Fig. 1,C). The same signals also elicited the strong cytoplasmic accumulation of thymocyte HuR (Fig. 1,D), confirming observations on T cell lines (19). However, both TCR and apoptotic signals (CD95/Fas or corticosteroids) induced the cytoplasmic cleavage of HuR, which is known to induce apoptosome activation (Fig. 1 D; Ref 18).

The differential response of HuR to thymocyte signals suggested that it may be involved in the cellular events driving thymocyte maturation. To reveal HuR’s role in these events, we induced the Cre-mediated inactivation of its murine Elavl1 locus in thymocytes by crossing LckCre transgenic mice to mice bearing an inactivatable, loxP-containing allele (Elavl1^fl; Supplemental Fig. 1) (26). Because monoallelic and compound heterozygote mutant thymocytes expressed comparable levels of HuR, we present LckCre⁺Elavl1^fl/+ as controls for LckCre⁺Elavl1^fl/fl. The recombination of the Elav1l^fl locus and the loss of HuR protein were consistent among different LckCre⁺Elavl1^fl/fl thymi, whereas intracellular flow cytometry validated that >90% of LckCre⁺Elavl1^fl/fl TCRαβ⁺ thymocytes and peripheral T cells possessed a HuR-null (HuR⁻) phenotype (Supplemental Fig. 1). At the age of 4 wk, mutant thymi contained a normal number of DP and SP thymocytes but increased numbers of DN cells (Fig. 1, E and G); however, at 10 wk, cellular increases extended to all thymocytes in LckCre⁺ Elavl1^fl/fl mice and persisted past this age, thus contributing to delays in thymic involution (Fig. 1,G and not shown). In contrast, the peripheral T cell counts in these mice were progressively reduced to <50% of control values and were most prominently diminished in aged mice (Fig. 1, F and G and not shown). Thus, the loss of HuR in thymocytes disturbed the balanced representation of T cells in lymphoid organs, which could be due to aberrations in the staged progression of thymic T cell development.

Hypercellular thymi may result from the defective control of DN cells, which provide the thymocyte pool for selection. In LckCre⁺ Elavl1^fl/fl thymi, HuR⁻ cells were first detected between DN2 and DN3, coinciding to the emergence of intracellular TCRβ (Fig. 2,B). The median ratio of HuR⁺/HuR⁻ cells at DN3 in mutant mice was 1.6:1.0 but changed dramatically in favor of HuR⁻ cells at DN4 (1.0:4.0), accounting for increased DN4 counts and the predominance of HuR⁻ cells in descendant immature SP, DP, and mature SP thymocytes as well as peripheral T cells (Fig. 2, A and B and supplemental Fig. 2). The progressive prevalence of HuR⁻ T cells could be either due to the staged recombination of the Elavl1 locus or to the increased fitness of HuR⁻ cells. To determine which was the cause, we used competitive repopulation of the thymus in immunodeficient hosts to measure the fitness of mutant cells via their representation in each stage. CD45.1 bone marrow progenitors (competitor) were mixed either with CD45.2 LckCre⁺ Elavl1^fl/+ (control) or CD45.2 LckCre⁺ Elavl1^fl/fl (test) progenitors and transplanted onto irradiated CD45.2 RAG1^−/− mice. Injection of competitor with control progenitors resulted in the equal representation of descendant thymocyte subsets; in contrast, the mix of competitor with test progenitors was followed by the overrepresentation of LckCre⁺ Elavl1^fl/fl cells at DN4 and DP thymocytes (Fig. 2, C and D), suggesting that HuR controls the expansion of the pre-TCR-expressing immature thymocytes. To verify this, we assessed the proliferative competence of HuR⁻ thymocytes by giving a pulse of BrdU in vivo, followed by cell cycle analysis. LckCre⁺ Elavl1^fl/fl thymi showed a clear increase in the number of BrdU⁺ DN cells in the S or G2/M phases and their descendant DPs (Fig. 2,E). The control of DN′s cell cycle is known to be achieved via a checkpoint that allows for DNA repair and elimination of aberrant TCRβ rearrangements and requires the activation of the p53 tumor suppressor and cyclin-dependent kinase (Cdk) inhibitors like p21 and p27. HuR is known to bind to p53, p21, and p27 mRNAs (11, 12, 33), but their steady state levels remained unaltered in sorted mutant DN cells (not shown). Similarly, the levels of p21 and p27 proteins were not grossly affected by HuR’s loss; however, the p53 protein was reduced 3-fold in mutant DN cells (Fig. 2 F). Thus, HuR controls the cell cycle progression of DN thymocytes during TCRβ selection relating to its effects on the expression of the p53 suppressor.

Our reconstitution experiments revealed additional defects that could not be observed in a noncompetitive environment. In contrast to HuR⁻ DP thymocytes, their SP descendants were reduced in chimeric thymi and diminished in the periphery of reconstituted mice (Fig. 3, A and B), suggesting defects in the generation of mature SP cells via positive selection. Cells undergoing selection are marked by an increase in surface TCR and CD5 levels and the transient expression of the early activation marker, CD69. LckCre⁺ Elavl1^fl/fl DP cells contained a lesser number of HuR⁻ cells initiating positive selection (TCRβ^intCD69^high or CD5^high) and, particularly, of cells undergoing positive selection (TCRβ^highCD69^high or CD5^high) (Fig. 3, C and D). Furthermore, the proliferative response of LckCre⁺ Elavl1^fl/fl thymocytes to TCR agonists, but not mitogens, was reduced (Fig. 3,E), suggesting defects in TCR signaling. This was exemplified further by the impaired activation of intracellular signals in HuR⁻ thymocytes pulsed with TCR agonists. Although the protein levels of TCR adaptor and distal kinases like p56^lck, ZAP-70, and PKCθ remained unaltered in the absence of HuR, their phosphorylation was compromised in stimulated Elavl1^Cre/− thymocytes (Fig. 3 F). Thus, HuR modulates TCR signals driving thymocyte positive selection.

A defect in positive selection could account for the loss of peripheral HuR⁻ T cells but not for their presence as SP thymocytes. This paradox was also apparent in mice with TCR-transgenic thymocytes that recognize alloreactive MHC class I-restricted Ags like the F5-TCR that recognizes a viral Ag, or the HY-TCR recognizing the male autoantigen in female mice (supplemental Fig. 3). We note that HuR’s loss was partial in tgHY⁺ LckCre⁺ Elavl1^fl/fl T cells, probably due to the surface expression of the HY-TCR in DN cells enforcing selection at the time of recombination of the Elavl1 locus. Still, both systems suggested that HuR⁻ SP cells accumulate in the thymus due to their defective egress. To test this, we labeled thymocytes via the intrathymic injection of FITC and analyzed peripheral naive (CD44^low) T cells for FITC⁺ recent thymic emigrants (RTEs). In contrast to control groups, RTE counts were reduced in the periphery of LckCre⁺ Elavl1^fl/fl mice (Fig. 4, A and B). In compliance with an impairment in egress, HuR⁻ SP thymocytes were enriched in postselection TCRβ⁺CD69^lowCD62L^high cells (Fig. 4,C) that respond to chemotactic signals for their intrathymic movements and accumulate in the presence of drugs inhibiting these processes (34). Such signals include those transduced by the shingosine-1-phosphate receptor 1 (S1P₁/EDG-1) acting through the Kruppel-like transcription factor 2, as well as the chemokine receptors CXCR4 and CCR7 (3, 4). The expression of these molecules was not altered in HuR⁻ cells (supplemental Fig. 4) and the migration of HuR⁻ and HuR⁺ SP thymocytes to S1P was comparable (not shown). In contrast, HuR⁻ SP thymocytes migrated poorly toward the CXCR4 ligand, CXCL12/SDF1, or the CCR7 ligand, CCL21 (Fig. 4,D). Furthermore, these chemokines changed the subcellular distribution, but not the abundance of HuR in proficient thymocytes (supplemental Fig. 4). A prevalence of nuclear HuR was observed in CCL21-treated thymocytes, whereas the exact opposite was observed after SDF-1 treatment (Fig. 4 E), suggesting that chemokines control HuR’s functions differentially, for the optimal migration and egress of postselection SP thymocytes.

The defective egress of HuR⁻ SP thymocytes could also mask defects in thymocyte deletion, which is the basis of negative selection. This was suggested by the refractory response of HuR⁻ DP thymocytes to anti-CD3-induced deletion in vivo and their defective ex vivo apoptotic response to anti-CD3 in the presence of anti-CD28 but not corticosteroids (Fig. 5, A and B and supplemental Fig. 5). Similarly, HuR’s loss hindered Ag-induced deletion, as demonstrated by the poor elimination of F5-TCR⁺HuR⁻ thymocytes exposed to a chronic 4-day administration regime of their cognitive NP68 peptide (Fig. 5,C and supplemental Fig. 5). To extend our observations to autoantigen-induced deletion, we analyzed the elimination of autoreactive HY-TCR⁺ transgenic thymocytes, which occurs in male mice. We hypothesized that because HuR⁻ HY-TCR⁺ thymocytes were partially present in tgHY⁺ LckCre⁺ Elavl1^fl/fl female mice, presumptive defects in their deletion and egress should enforce their presence in the autoreactive male environment. In control males, HY-TCR⁺ DP thymocytes were consistently deleted, whereas remnant CD8⁺ SPs possessed an anergic HY-TCR^low phenotype. Contrastingly, and despite the early loss of DP thymocytes in tgHY⁺ LckCre⁺ Elavl1^fl/fl thymi, a HuR⁻HY-TCR^high CD8⁺ SP subset emerged at 5 wk of age and accounted for a 50% increase in total CD8⁺ SP thymocytes by 15 wk of age (Fig. 5, D and E and supplemental Fig. 5). Strikingly, peripheral CD8⁺ HuR⁻ HY-TCR^high subsets could not be detected in tgHY⁺ LckCre⁺ Elavl1^fl/fl mice (Fig. 5 E), demonstrating that the loss of HuR hinders the deletion of autoreactive thymocytes but does not allow their exiting from the thymus.

Thymocyte deletion is influenced by proapoptotic signals emanating from death receptors like Fas (CD95/Apo-1), which supports clonal deletion at high-Ag concentrations (35). The expression of Fas was not altered in LckCre⁺ Elavl1^fl/fl thymocytes (supplemental Fig. 5), but these cells showed a minimal apoptotic response to anti-Fas ligation and a consistent lack of cleaved caspases-8 and -3 (Fig. 5, F and G). Collectively, HuR is required for Ag and death-receptor induced thymocyte deletion.

To identify gene networks affected by HuR’s dysfunction, we performed microarray hybridizations using RNA from control and HuR⁻ thymocytes stimulated for 0, 4, and 12 h by mitogens. Two hundred forty one unique RNAs were identified as differentially up (140) or down (101) in HuR⁻ thymocytes (Supplemental Table I, a and b). Some of the previously reported HuR-target mRNAs (e.g., cyclins, cytokines, and transcription factors) were expressed in thymocytes, but appeared unaffected by HuR’s loss (Supplemental Fig. 6). To relate the expression data to the phenotypes of HuR⁻ thymocytes, we used the Endeavour relational software to perform a computational prioritization of differentially expressed genes (Table I) based on a training set of 177 genes (supplemental Tables) controlling thymic cellularity. The hyperproliferation of HuR⁻ DN thymocytes, related to the increased expression of Cdc25A/B phosphatases which promote cell cycle progression. Defects in antigenic signaling related to increases in mRNAs encoding the TCR constituents and downstream signalers (CD2, CD3ε, Vav1, Act3, Runx1) but also to decreases in the mRNA of the Src-like adaptor protein Sla, which inhibits CD3ζ-signaling in DP cells and the Toe-1 mRNA, which is a target of the Egr-1 transcription factor required for positive selection. Most strikingly, the defective deletion of HuR⁻ thymocytes related to changes in members of the TNF/TNFR family. Subsequent analyses validated the direct involvement of HuR in the modulation of such molecules. The biosynthesis of TNF-α was augmented in activated HuR⁻ thymocytes, whereas its mRNA was associated with thymocyte HuR in R-IP assays (Fig. 5, H and I and supplemental Fig. 5), confirming previous observations on HuR’s negative control over TNF biosynthesis (10). In light of TNF’s positive role in thymocyte deletion (36), and the invariable expression of its receptors in HuR⁻ thymocytes (supplemental Fig. 5 and not shown), we postulate that HuR affects downstream TNFR signaling as in the case of Fas. In contrast, however, to TNFRs and Fas, the expression of TNFRSF25/DR3, which promotes negative selection (37), was reduced in HuR⁻ thymocytes and particularly in DP cells. Most importantly, the DR3 mRNA precipitated with thymocyte HuR in HuR-R-IP assays (Fig. 5, H and I and supplemental Fig. 5), suggesting that HuR can control both proximal and distal proapoptotic signals. Collectively, our profiling data connect HuR’s control over thymocyte maturation to gene expression programs controlling cell cycle, TCR, and TNFR/death receptor signaling.

In this report, we revealed HuR’s pleiotropic control over TCRαβ⁺ T cell development via its genetic ablation in thymocyte precursors. On the one end, HuR’s loss caused an increase in thymic cellularity due to hyperproliferating DN cells, nondeleted thymocytes, and the nonmigrating mature T cells. On the other end, the loss of peripheral HuR⁻ T cells was attributed to their defective positive selection and the defective egress of remnant cells. These aberrations reflected HuR’s involvement in thymocyte intrinsic/extrinsic signals that could only be revealed in an in vivo multiparametric context.

The hyperproliferation of HuR⁻ DN thymocytes highlighted HuR’s role as a cell cycle suppressor of pre-T cells undergoing β-selection. HuR’s involvement in cell-cycle control has been inferred previously via its affinity for mRNAs encoding cyclins, tumor suppressors, proto-oncogenes, Cdk inhibitors, and enzymes promoting DNA repair, as well as apoptotic modulators (15, 38). In our study, the increased proliferation of HuR⁻ DN cells related to their reduced content in one of HuR’s known targets, p53, which plays a dominant role toward the growth arrest of pre-T cells undergoing β-selection. In the absence of a functional pre-TCR, p53 accumulates to aid cell clearance by apoptosis, whereas productive TCRβ rearrangements inactivate p53, resulting in cell cycle entry. p53^null DN thymocytes escape this checkpoint and become DP even in the absence of a functional pre-TCR (39). This role for p53 may also apply for HuR⁻ thymocytes, as suggested by the increased expression of Cdc25A and B phosphatases, which promote G1/S-phase transition and activation of the mitotic Cdk1/cyclin B complex. Similar changes in Cdc25 expression have been reported in p53^null cells (40) and p53 is known to repress the transcription of Cdc25 genes (41, 42). Consistent with HuR’s role in promoting p53 mRNA translation (12), we propose that HuR may enhance the expression of p53 to promote cell cycle arrest in DN thymocytes. However, the reduction of p53 protein in HuR⁻ DN thymocytes may also result from alterations in its posttranslational modifications controlling its stability (43); this points toward HuR’s alternative involvement in signals guiding p53’s modification and cell cycle control. Consistent with this suggestion, prior data demonstrated that HuR can be phosphorylated by the checkpoint kinase Chk2 to bind to its targets during oxidative stress (15) and that Chk2 contributes to the stabilization of p53 (44). Conversely, HuR functions can be blocked by its nuclear retention following phosphorylation by Cdk1 (16), whereas Cdk1 is inhibited by p53 and p21 (45).

The prevalence of HuR⁻ DN cells was counteracted by the lack of TCR signals needed for positive selection. Because antigenic/TCR signals alter the abundance and localization of HuR in T cells, we postulate that they converge to HuR to control its functions for the prudent execution of gene expression programs driving thymocyte activation during positive selection. This supposition is supported further by the efficacy of a small molecule inhibitor of HuR’s shuttling that blocks T cell activation in vitro (46).

Alternatively, HuR may also affect the expression of TCR signalers, as suggested by the loss of proximal TCR signaling in HuR⁻ cells. Our profiling data provided important clues for HuR’s interference to TCR signaling by the altered expression of TCR constituents or their regulators, like the CD3ε-associated molecule Sla in HuR⁻ cells, which is known to block the signaling functions of CD3ζ (47). Supporting our hypothesis is a parallel study identifying mRNA targets of HuR that change during T cell activation in Jurkat cells, and encode proteins imminent to TCR signaling (N. Mukherjee and J. D. Keene, submitted for publication).

HuR’s requirement in TCR signaling was substantiated further by the defective deletion of HuR⁻ thymocytes, because a reduction in the thresholds of (auto)antigenic signaling could account for this effect. However, HuR is also involved in extrinsic, proapoptotic, cytokine signals such as those induced by Fas or other TNFR family members like TNFRI or DR3. Thus, HuR can modulate proapoptotic cytokine signaling in two ways. It can bind and affect the expression of apoptotic ligands (e.g., TNF) or death receptors (e.g., DR3). Adding to our data, a recent biochemical report showed that HuR can promote the alternative splicing of Fas pre-RNA to encode for its transmembrane rather than its soluble isoform (21). Although HuR⁻ thymocytes expressed normal levels of intracellular and transmembrane Fas protein, we cannot exclude the possible overproduction of soluble Fas as an additional mechanism affecting in vivo thymocyte deletion. On top of its biosynthetic effect on ligands/receptors, HuR’s involvement in proapoptotic signals can be inferred by the sharing of signal transducers from DR3, Fas, and TNFRI (48, 49). How can HuR interfere with these signals? A recent report (18) demonstrated that, during lethal challenge, HuR is cleaved at an aspartate residue by a mechanism requiring caspase-7 and -3. Cleavage releases HuR from its translocation partner pp32/PHAP-I, which then acts as a proapoptotic factor by stimulating the activity of the apoptosome complex. In our study, cleaved HuR was detected in TCR- or Fas-stimulated thymocytes, suggesting that its absence may block the activation of the apoptosome during thymocyte deletion. However, HuR’s cleavage was also induced during corticosteroid-induced thymocyte deletion, which was not affected by HuR^’s loss. Furthermore, the refractory response of HuR⁻ thymocytes to Fas related to the defective activation of upstream caspase-8. Thus, it seems that HuR controls both upstream proapoptotic signals as well as the activation of the apoptosome.

The loss of HuR’s control over thymocyte egress caused the accumulation of postselection HuR⁻ SP thymocytes and the reduction in peripheral HuR⁻ RTEs. Two chemokine-dependant processes have been linked to egress of mature thymocytes: 1) their CCR7- or CXCR4-dependent retention/repulsion toward the medullary endothelial cells, and 2) their S1PR-dependent attraction toward the peripheral blood. HuR⁻ SP thymocytes showed a compromised response only to CXCR4 and CCR7 ligands, suggesting that HuR controls medullary movements and not thymocyte attraction to the peripheral blood. This could also provide an explanation as to why egress is not fully blocked as in cases of S1PR blockade (50). Similarly, the differential cytoplasmic response of HuR to CXCR4 and CCR7 ligands may reflect its differential involvement in cellular retention vs repulsion.

At the molecular level, we postulate that HuR acts both as a regulator of gene expression and as a modulator of signal transduction. Given the diversity of T cell phenotypes and responses, a subset/signal restricted analysis on proteomic/ribonomic interactions needs to be performed to reveal the complete HuR “interact-ome” in T cells. Clearly however, our data are against the simplistic dogma where HuR acts to control the use of mRNAs in a common manner because the limited number of HuR-interacting mRNAs examined herein show variable responses to HuR’s absence. We note however the observed changes in these mRNAs may either reflect differences in mRNA stability/translation relating to direct HuR functions or extent to other direct/indirect transcriptional/posttranscriptional processes including the functions of other RBPs.

Finally, and in the context of adaptive immunity, many of the thymocytic signals affected by HuR’s dysfunction are also implicated in peripheral T cell responses under physiological or inflammatory settings. In that context, the peripheral T cell loss observed in mutant mice could also results from defects in peripheral subset composition, activation, proliferation, and survival. Similarly, and based on our findings on the control of thymocyte deletion of autoreactive thymocytes, HuR could affect negative selection as well as peripheral tolerance to auto- or innocuous Ags and autoimmunity. These important immunological processes will be examined in the future, along with the combined effects of HuR, other RBPs, and noncoding regulatory RNAs on adaptive immunity.

Thanks to V. Katsanou for her contributions at the early stages of this project; P. J. Lager for his support in the microarray experiments; A. Potocnik for training on intrathymic FITC injections, CD45.1 congenic strains, and comments; D. Kioussis, B. Malissen, and G. Kollias for mice; and G. Kassiotis, C. Mamalaki, M. Apostolaki, and D. Graf for critical discussions and comments.

The authors have no financial conflict of interest.