GATA-3 Regulates the Homeostasis and Activation of CD8⁺ T Cells

GATA-3, the third member of the C2C2-type zinc finger GATA family of transcription factors, was first cloned by its ability to bind a T cell–specific enhancer located in the TCR α gene (1, 2). Although GATA-3 is expressed in several types of cells, its expression is limited to cells of T lineage among hematopoietic cells (3). During the development of thymocytes, the expression of GATA-3 is induced after β-selection and MHC-mediated positive selection (4). The level of GATA-3 remains high in CD4 single-positive thymocytes but is substantially reduced in CD8 single positive thymocytes. Among the functional subsets of peripheral Th cells, GATA-3 is highly enriched in Th2 cells, and its expression has been used to define this subset of Th cells.

The function of GATA-3 in the development and differentiation of thymocytes and Th cells have been well characterized. GATA-3–deficient (G3 knockout [KO]) ES cells or hematopoietic stem cells failed to contribute to the T lineage in RAG-deficient or irradiated mice (5, 6), indicating that GATA-3 is indispensable for the thymic seeding or differentiation of early thymic progenitors. In the presence of Notch signals, the level of GATA-3 has to be “just right” to ensure the survival of DN1 and DN2 cells (7). Deficiency of GATA-3 at late DN2 stage results in a profound block in β-selection because of impaired production, but not rearrangement, of TCRβ-chain (8). Deficiency of GATA-3 at early double-positive (DP) stage, however, resulted in a virtual absence of CD4SP thymocytes despite a near normal number of CD8SP cells (8, 9). How GATA-3 regulates the development of CD4SP is still largely unknown. Early TCR signaling cascades were apparently intact in G3KO DP cells. Examination of positive selection of DP cells with various MHC class II–restricted TCR transgene found either no or only a modest defect in the absence of GATA-3. It has been shown that GATA-3 directly regulates the expression of Th-POK (9), a transcription factor that is also essential for the development of CD4SP cells (10, 11). However, a Th-POK transgene was unable to rescue the development of GATA-deficient CD4SP cells.

GATA-3 is also a critical transcription factor of Th2 cells (12, 13). It is preferentially expressed in Th2 cells compared with Th1 cells (14). Its expression in naive Th cells is induced by IL-4/Stat6 (15, 16), IL-2/Stat5 (17), and Notch signal pathways (18), all of which are important for the differentiation of Th2 cells. It can function as a chromatin-remodeling factor of the IL-4/IL-5/IL-13 locus as well as a direct transcription factor of IL-5 and IL-13 genes. It can also inhibit the activity of Stat4 and Runx3, thereby inhibiting Th1 transcription program (19). In agreement with these in vitro data, mice rendered deficient in GATA-3 only in mature T cells were unable to mount type 2 Th immune responses and were highly susceptible to parasitic infection (12, 13).

GATA-3 is expressed, albeit at a low level, in CD8SP thymocytes and CD8⁺ peripheral T cells (4, 20). Although deficiency of GATA-3 had little impact on the absolute number of polyclonal CD8SP thymocytes, G3KO CD8⁺ peripheral T cells also displayed impaired production of type 2 cytokines even under optimal culture conditions (12). Despite these observations, the role of GATA-3 in regulating the development, homeostasis, activation, and effector function of CD8⁺ T cells is still poorly characterized.

In this report, we studied mice (G3KO), in which the deletion of GATA-3 was mediated by a CD4cre transgene. We show that G3KO CD8SP polyclonal thymocytes displayed a defect in downregulation of CD24. In addition, G3KO mice developed age-dependent lymphadenopathy because of abnormal expansion of CD8⁺ T cells mediated by a cell-extrinsic mechanism. Paradoxically, G3KO CD8⁺ T cells were less efficient than control cells when undergoing homeostatic expansion in lymphopenic environments. Activated G3KO CD8⁺ T cells were also less efficient in killing Ag-bearing tumor cells in vivo. The impaired homeostatic expansion and in vivo Ag-specific killing was partly attributed to hyporesponsiveness to Ag stimulation. We further demonstrate that GATA-3 was essential for sustaining, but not initiating, activation signals after TCR engagement.

T cell–specific G3KO mice have been described previously (8). OT-1 mice were obtained from The Jackson Laboratory and crossed with G3KO mice. The animals were housed under specific pathogen-free conditions, and all animal experiments were performed in accordance with the institutional guidelines for animal care at Dana-Farber Cancer Institute under approved protocols. Male or female mice aged 4–12 wk were used. In all experiments, littermates (Floxed/Floxed [FF]) carrying the floxed GATA-3 allele, but not the CD4cre transgene, were used as controls.

Naive (CD62L^hiCD44^lo) CD8⁺ T cells were sorted from lymph nodes and spleens and activated with plate-bound anti-CD3 (ranging from 1 to 10 μg/ml) and soluble anti-CD28 (2 μg/ml) in the presence or absence of human IL-2 (100 U/ml). In some experiments, sorted naive CD8⁺ T cells were first stained with CFSE (Invitrogen, Carlsbad, CA) prior to stimulation.

The following clones of Ab were purchased from BioLegend (San Diego, CA) and used for cell surface staining: CD4(RM4-5), CD8(53-6.7), TCRβ(H57-597), L-selectin (CD62L[MEL-14]), CD69(H1.2F3), CD25(PC61), CD24(M1/69), CD45.2(104), CD44(IM7), Qa2(695H1-9-9), KLRG1(2F1/KLRG1) Vβ5(MR9-4), and Fas ligand (FasL[MFL3]). Flow cytometry was performed on a FACSCanto or FACSCanto II and analyzed with FlowJo software (Tree Star, Ashland, OR).

In each sample, 5 × 10⁵ naive CD8⁺ T cells, either unstimulated or stimulated, were lysed in freshly prepared radioimmunoprecipitation assay buffer containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM PMSF, 1 mM EDTA, 1% Triton X-100, 1% sodium deoxycholate, and 0.1% SDS. Cell lysate was separated from debris by centrifugation at 12,000 rpm for 10 min. Lysate was loaded onto 9% polyacrylamide gels and transferred onto polyvinylidene difluoride membrane (Polyscreen; PerkinElmer). The membrane was subsequently blocked in 5% milk and probed with Abs according to the manufacturer’s protocol. The following Abs were used. Anti-ERK2 (C-14), anti-JNK (F-3), anti-p38α (C-20), anti-Lat (11B.12), and anti-Hsp90 were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Anti–phospho-p44/42 MAPK (Erk1/2) (Thr²⁰²/Tyr²⁰⁴), anti–phospho-SAPK/JNK (Thr¹⁸³/Tyr¹⁸⁵), anti–phospho-p38 MAPK (Thr¹⁸⁰/Tyr¹⁸²), and anti–p-Zap70(Y319) were purchased from Cell Signaling Technology (Danvers, MA). Anti-Zap70 (1E7.2) was purchased from BioLegend. Proteins were visualized using an ECL kit (PerkinElmer).

Total RNA was purified using a TRIzol Plus kit (Invitrogen). First-strand cDNA synthesis was performed with 200 ng total RNA using the QuantiTect Reverse Transcription kit (Qiagen, Hilden, Germany). Gene expression levels were determined by real-time PCR analysis performed using the Brilliant SYBR Green QPCR kit, according to the manufacturer’s protocol (Stratagene, La Jolla, CA), on a MX-3000P apparatus (Stratagene) using the following cycling conditions: denaturation at 95°C for 30 s, annealing at 56°C for 60 s, and extension at 72°C for 30 s. Primer sets were designed using the Primer3 web utility (http://frodo.wi.mit.edu/). Levels of mRNA were adjusted for differences in β-actin expression. The following primers are used: c-fos (5′-GTAGAGCAGCTATCTCCTGA; 3′-ACGCAGACTTCTCATCTTC); c-Jun (5′-ACAGCTTAAGCAGAAAGTCA; 3′-CAACCAGTCAAGTTCTCAAG); perforin (5′-GCTGAGAAGACCTATCAGGA; 3′-TAGGAGGAGATGAGCCTGT); actin primer (5′-GGCTGTATTCCCCTCCATCG; 3′-CCAGTTGGTAACAATGCCATGT); and IL-2 primer (5′-AGCAGCTGTTGATGGACCTA; 3′-CGCAGAGGTCCAAGTTCAT).

Sandwich ELISA was performed using the following mAb pairs (BD Biosciences, Franklin Lakes, NJ): anti–IL-2 (JES6-1A12)/biotin- anti–IL-2 (JES6-5H4), and anti–IFN-γ (R4-6A2)/biotin-anti–IFN-γ (XMG1.2).

OVA-transfected B16-derived melanoma cell line B16-OVA cells were provided by Dr. S. Gerber (University of Rochester Medical Center, Rochester, NY). B16-OVA cells were cultured in DMEM supplemented with 10% FCS and penicillin/streptomycin, trypsinized, and washed twice in PBS. A total of 1 × 10⁵ cells were injected s.c. into shaved back of C57BL/6 mouse. Four days later, host animals were injected i.v. with 7 × 10⁶ FF/OT-1 or G3KO/OT-1 CD8⁺ T cells that were prestimulated with SIINFEKL peptide derived from OVA (5 μg/ml; Invitrogen) in the presence of 100 U/ml IL-2 for 3 d. Host mice were sacrificed at day 16, and the tumor diameter was determined as the mean of the largest diameter and the diameter at right angle. Tumors were collected and grinded by cover slide. After passing through 70-μm filters, cells were recovered and stained with surface Abs, collected through a FACS canto II, and analyzed by FlowJo software.

Student t test was used for statistical analysis: *p < 0.05; **p < 0.005; ***p < 0.0005.

After positive selection and lineage commitment, both CD4SP and CD8SP thymocytes downregulate CD24 prior to emigrating out of the thymus. We have previously shown that CD4cre-mediated deficiency of GATA-3 had little impact on the development of polyclonal CD8SP thymocytes despite the virtual absence of CD4SP cells. Although the total number of TCR^hi CD8SP thymocytes was normal, we found that these mice contained significantly fewer mature TCR^hiCD24^lo CD8SP thymocytes and reciprocally more immature TCR^hiCD24^hi CD8SP cells (Fig. 1A). The percentage of CD24^lo cells among CD8SP subset was ∼45% in FF mice but was only 25% in KO mice. Despite the defect in downregulating CD24, G3KO CD8SP thymocytes did express a normal level of Qa2, another marker of mature SP thymocytes (Fig. 1A, right lower panel) (21). These data suggest that GATA-3 may be required for the final maturation step of polyclonal CD8SP thymocytes. To further examine the impact of GATA-3 deficiency on the development of CD8 lineage, we introduced OT-1 TCR transgene into G3KO mice. We found that the thymus of G3KO/OT-1 mice contained ∼40% more cells than FF/OT-1 mice (data not shown). Eighty percent of G3KO/OT-1 thymocytes, compared with 40% of FF/OT-1 thymocytes, were DP (CD8^hiVβ5^lo) cells, resulting in an ∼4-fold increase in the number of DP cells in G3KO/OT-1 mice (Fig. 1B). Despite the marked increase in the number of DP cells, there was a substantial reduction in the number of postselected (CD8^loVβ5^hi) and CD8SP (CD8^hiVβ5^hi) thymocytes in G3KO/OT-1 mice. The residual postselected G3KO/OT-1 thymocytes also did not upregulate CD69 (Fig. 1C), an indirect readout of positive selection. The CD69 mean fluorescence intensity (MFI) in both FF/OT-1 and G3KO/OT-1 DP populations was very comparable (∼330). The CD69 MFI in FF/OT-1 postselected cells expectedly increased to ∼1184, whereas the CD69 MFI of G3KO/OT-1 postselected cells actually decreased to ∼168. These observations indicate that deficiency of GATA-3 results in a profound defect in the positive selection of OT-1 thymocytes. The remaining G3KO/OT-1 CD8SP thymocytes that survived positive selection still displayed a defect in the downregulating CD24 (Fig. 1D).

Despite the reduction in thymic output of mature CD8⁺ T cells and virtual absence of CD4⁺ T cells, the absolute number of peripheral CD8⁺ T cells of G3KO mice were comparable to that of control mice up to 4 wk of age. Thereafter, G3KO mice gradually developed lymphadenopathy (Fig. 2A), which peaked at ∼6 wk after birth. At that age, the number of total cells per inguinal lymph node of G3KO mice were ∼3-fold more than FF mice and were made up of mainly CD8⁺ T and B cells (Fig. 2B). Although ∼60% of FF lymph node CD8⁺ T cells displayed a naive (CD44⁻CD62L⁺) phenotype, only 20% of G3KO CD8⁺ T cells were naive cells. Reciprocally 50% of G3KO CD8⁺ T cells, compared with 20% of FF CD8⁺ T cells, were CD44⁺CD62L⁻ “memory/effector” cells (Fig. 2C, left two panels). Thus, the abnormal expansion of CD8⁺ T cells was mainly due to memory/effector cells, whereas the number of naive CD8⁺ T cells in lymph nodes was very comparable between FF and G3KO mice (Fig. 2C, right two panels). Interestingly, the lymphadenopathy and abnormal expansions of CD8⁺ T and B cells became less obvious 9 wk after birth and almost fully resolved when the mice reached 12 wk of age. The percentage of memory and naive CD8⁺ T cells also nearly returned to normal. We found that 12-wk-old G3KO mice still had very few CD4⁺ T cells and that their CD8⁺ T cells still had an undetectable level of GATA-3 protein (data not shown). Therefore, the resolution of lymphadenopathy was not due to repopulation by GATA-3–sufficient T cells. The lymphadenopathy and abnormal expansion of CD8⁺ T was not driven by autoantigens because this phenotype is even more striking in G3KO/OT-1 mice (Fig. 2D). The number of CD8⁺ T cells and B cells in lymph nodes of G3KO/OT-1 mice was almost seven times of that of FF/OT-1 mice. More than 35% of G3KO/OT-1 CD8⁺ T cells, compared with <10% of FF/OT-1 cells, displayed memory/effector features. In contrast to G3KO mice, the numbers of both naive and memory/effector CD8⁺ T cells in lymph nodes were increased in G3KO/OT-1 cells (Fig. 2D, right panel), indicating that the abnormal expansion is not limited to memory CD8⁺ T cells. G3KO mice remained relatively healthy without any evidence of autoimmune features in specific pathogen-free environment up to 12 mo of age despite the age-dependent lymphadenopathy.

The abnormal expansion of CD8⁺ T cells observed in G3KO mice could be caused by a cell-intrinsic defect. The deletion of GATA-3 gene in G3KO mice is mediated by a CD4-driven Cre, which can be expressed in some non-T cells, including dendritic cells and lymphoid tissue-inducer cells. Some of these cells may also express a low level of GATA-3. Thus, the abnormal expansion of CD8⁺ T cells could also be mediated by a cell-extrinsic defect. To distinguish these two scenarios, we labeled wild-type (WT) congenic CD45.1 CD8⁺ T cells with CFSE and transferred the cells into 4-wk-old G3KO or FF mice. Host mice were examined 2 wk later. Four-week-old mice were chosen because the lymphadenopathy has yet to develop in this age. Most of the donor cells (>80%) recovered from peripheral lymph nodes of FF host mice remained undivided (Fig. 3A). In contrast, ∼60% of the donor cells recovered from G3KO host mice had undergone at least one division, and their number per inguinal lymph node was twice of their counterparts (Fig. 3A; data not shown). These data strongly suggest that the abnormal expansion of CD8⁺ T cells in G3KO mice was driven by a CD8⁺ T cell–extrinsic mechanism.

In a reciprocal experiment, we purified naive G3KO or FF CD8⁺ T cells, labeled them with CFSE and adoptively transferred them into RAG2KO mice, allowing the cells to undergo homeostatic expansion. We then recovered donor cells from lymphoid organs of host animals four days after cell transfer. We found that FF donor cells constituted ∼0.6% of lymph node cells (Fig. 3B). More than 75% of the FF donor cells had undergone at least one round of division. We were surprised to find that <0.2% of lymph node cells from host animals that received G3KO CD8⁺ T cells were donor cells. Only 40% of those G3KO CD8⁺ T cells were proliferating (Fig. 3B). These data are consistent with the notion that the abnormal expansion of CD8⁺ T cells seen in G3KO mice was not mediated by a T cell–intrinsic mechanism. Rather, G3KO CD8⁺ T cells demonstrated a cell-intrinsic defect in undergoing homeostatic expansion.

One possible explanation for the paradoxical result is that G3KO memory/effector cells expand more robustly than control cells through a cell-intrinsic mechanism. In disagreement with this explanation, we found that G3KO memory/effector (CD44⁺CD62L⁻) CD8⁺ T cells also proliferated poorly when transferred into RAG2KO mice compared with their FF counterpart (Fig. 3C). Thus, G3KO memory CD8⁺ T cells also have a cell-intrinsic defect in undergoing homeostatic expansion.

The impaired homeostatic expansion was also observed with G3KO mature CD8SP thymocytes and was even more striking in the presence of competition. We isolated TCR^hiCD24^lo CD8SP thymocytes from G3KO and congenic (CD45.1) WT mice, mixed them at 1:1 ratio, labeled them with CFSE, and transferred them into RAG2KO mice. When we examined the host animals 4 d later, we found virtually no G3KO cells in lymph nodes of host animals despite the presence and expansion of WT donor cells (Fig. 3D). The KO/WT ratio was reduced to <0.01 and declined further 5 wk after cell transfer. The reduced number of G3KO cells in lymph nodes of host animals was not due to abnormal distribution of G3KO donor cells to other lymphoid organs because we detected very few donor cells, either G3KO or control, in spleen and peripheral blood of these animals (data not shown).

Homeostatic expansion of T cells requires tonic stimulation through TCR and cytokines, such as IL-7 and IL-15. We found that G3KO CD8⁺ T cells expressed a normal level of CD127, the α subunit of IL-7R (data not shown). To examine their responsiveness to IL-7, we cultivated naive G3KO and FF CD8⁺ T cells in vitro in the absence or presence of IL-7 for 4 d. In the absence of IL-7, only ∼1–2% of cells remained alive (Fig. 4A). The percentage of live cells increased to ∼25% in the presence of IL-7 and both G3KO and FF cells responded comparably. Similarly, G3KO CD8⁺ T cells responded normally to IL-15 stimulation.

We then examined whether G3KO CD8⁺ T cells were hyporesponsive to TCR stimulation. We stimulated naive G3KO and FF cells in vitro with anti-CD3 and anti-CD28. FF cells responded to anti-CD3 stimulation in a dose-dependent manner (Fig. 4B). However, the response of G3KO CD8⁺ T cells was markedly attenuated. Very few live G3KO CD8⁺ T cells were recovered after 4 d of stimulation, and the remaining live cells expressed a low level of activation markers, such as CD25 and CD69, and proliferated poorly (Fig. 4C; data not shown). Only ∼25% of live G3KO T cells were CFSE^loCD25^hi, whereas ∼90% of live FF cells were activated and proliferating. In addition, very little IL-2 or IFN-γ was detected in supernatant of G3KO cells (Fig. 4D). G3KO CD8⁺ T cells were also hyporesponsive to stimulation in an Ag-specific manner. We stimulated naive G3KO and FF OT-1 cells with ova peptide in the presence of autologous APC. Again, we found that G3KO OT-1 cells also responded less robustly to ova peptide (Supplemental Fig. 1A). These results collectively indicate that deficiency of GATA-3 has a global impact on the activation/proliferation of CD8⁺ T cells.

This hyporesponsiveness could be partly corrected with exogenous IL-2. When stimulated with higher doses of anti-CD3 (5 μg/ml or higher) in the presence of IL-2, the percentage of live cells based on a forward light scatter (FSC)/SSC gate was near normal in the absence of GATA-3 (Fig. 4E, upper left panel). However, the percentage of activated/dividing (CFSE^loCD25^hi) cells and the production of IFN-γ and IL-2 were still much lower in G3KO populations (Fig. 4E, other three panels). One possible explanation for the partial rescue by exogenous IL-2 is that G3KO CD8⁺ T cells have an intrinsic defect in producing IL-2. This scenario is unlikely because the production of IL-2 by comparably activated G3KO CD8⁺ T cells (stimulated with 5 μg/ml anti-CD3 in the presence of exogenous IL-2) and FF cells (stimulated with 1 μg/ml anti-CD3 in the presence of exogenous IL-2) was very similar. In addition, G3KO CD8⁺ T cells were capable of maintaining the expression of CD25 in the presence of IL-2 after withdrawal from CD3 stimulation, suggesting that deficiency of GATA3 does not affect IL-2/STAT5 signaling (Supplemental Fig. 1B).

One logical explanation for the hypo-responsiveness of G3KO CD8⁺ T cells is that GATA-3 regulates the strength of TCR signals. We therefore examined the TCR signaling cascade in CD8⁺ T cells stimulated with 2 μg/ml anti-CD3 and 2 μg/ml anti-CD28 in the absence of exogenous IL-2, a condition that sufficiently activated FF, but not G3KO, cells. Surprisingly, we detected no or little impairment in the level and kinetics of tyrosine phosphorylation within 10 min and phosphorylation of Jnk and p38 within 30 min after stimulation (Fig. 5A). NFAT phosphorylation/translocation and NF-κB nuclear translocation also progressed normally up to 4 h after stimulation (Fig. 5B). We did detect ∼30% reduction in the level of p-Erk in G3KO cells (Fig. 5A, 5C, Supplemental Fig. 2). This modest reduction in the level of p-Erk did not substantially affect the degradation of Bim, a known substrate of Erk (Fig. 5C) (22), or induction of IL-2, c-Fos, and c-Jun within the first 3 h after stimulation (Fig. 6A). In addition, the earliest features of activation, including shedding of CD62L, internalization of TCR, and downregulation of CD8 were undisturbed in the absence of GATA-3 (Fig. 6B). These data indicated that G3KO CD8⁺ T cells were capable of initiating activation signals in response to stimulation. These observations also prompted us to look at signaling events beyond 4 h after engagement of TCR. The initial wave of tyrosine phosphorylation died down by 4 h in both FF and G3KO cells (Fig. 5C). Subsequently, there was a resurgence of tyrosine phosphorylation in FF cells, starting 12 h after stimulation and propagating up to 24 h after stimulation. The level of p-Erk, p-Jnk, and p-p38, after the initial spike, dropped back to or below baseline by 4 h and was either maintained or slightly enhanced in FF cells 24 h after stimulation (Fig. 5C, Supplemental Fig. 2). However, the “second wave” of tyrosine phosphorylation was markedly attenuated in the absence of GATA-3, and G3KO cells were unable to maintain the level of p-Erk, p-Jnk, and p-p38 (Fig. 5D). Accordingly, more apoptotic cells were detected in G3KO population (Fig. 5E).

In agreement with the late signaling defects, G3KO CD8⁺ T cells were unable to sustain the expression of IL-2 after 4 h (Fig. 6A). In addition, the induction of CD69, CD25, and CD44 that normally takes place after 4 h of stimulation was marked reduced or did not occur (Fig. 6B). In these analyses, we gated only on “live” cells seen in FSC/SSC plots. More than 85% of the FF and G3KO “live” cells were negative for both 7-aminoactinomycin D (7AAD) and Annexin-V 24 h after stimulation (Supplemental Fig. 3). Therefore, the impaired expression of late activation markers in the absence of GATA-3 cannot be explained by contamination of apoptotic cells in G3KO population.

We then further investigated why G3KO CD8⁺ T cells could not maintain activation signals. As the levels of pTyr, p-Erk, p-Jnk, and p-p38 were markedly reduced in G3KO CD8⁺ T cells after the fourth hour, we postulated that deficiency of GATA-3 had a global impact on signaling in CD8⁺ T cells. Indeed, we found that the levels of several signaling molecules, although normal prior to stimulation, were much lower in G3KO cells than those in FF cells at the 12th hour or later time points (Fig. 5C). Stimulation through TCR is known to rapidly induce transient downregulation of Lck (23). This effect is mediated at the level of transcription and RNA stability (24). In agreement with the finding that G3KO CD8⁺ T cells were able to initiate activation signals, we found that the downregulation of Lck proceeded nearly normally in the absence of GATA-3. However, G3KO CD8⁺ T cells were unable to replenish Lck, a process started 12 h after stimulation in FF cells. In addition, G3KO CD8⁺ T cells could not maintain the level of Sos, Lat, and Zap70 (Fig. 5C, Supplemental Fig. 2). This reduction in the level of Lck, Sos, Lat, and Zap70 was not due to a global defect in gene expression because the level of Bcl2, Bim, and Erk was either normal or even slightly higher in G3KO CD8⁺ T cells at the same time points (Fig. 5C, Supplemental Fig. 2). Furthermore, the reappearance of CD62L, which occurred in FF cells 12 h after stimulation, took place slightly earlier and reproducibly reached a higher level in G3KO CD8⁺ T cells (Fig. 6B). Exogenous IL-2 partly restored the level of pTyr, Lck, and Sos in G3KO cells and modestly enhanced the level of CD25, a known target gene of STAT5, but still had no or little impact on the expression of CD69 and CD44 (Figs. 5C, 6B). Taken together, our data indicate that GATA-3 was required for maintaining the expression of a subset of signaling molecules and was therefore indispensable for the maintenance/progression, but not the initiation, of activation signals. Such a defect eventually may lead to hyporesponsiveness to Ag stimulation in G3KO CD8⁺ T cells.

Genome-wide search for GATA-3 binding sites in naive CD8⁺ T cells have been conducted using ChIP-seq technology (25). We searched the database for GATA-3 binding sites in the genes shown in Fig. 5C and Supplemental Fig. 2. We found that in vivo binding of GATA-3 within promoter or gene body did not predict whether the expression of a given gene would be lost in G3KO cells. For example, heavy GATA-3 binding was detected in the loci of Bcl2 and Bcl2l11 (encoding Bim), whereas no GATA-3 binding was detected in the Sos1 locus. Therefore, GATA-3 very likely maintained the expression of Lck, Sos, and Lat by both direct and indirect mechanisms.

We then set to examine whether deficiency of GATA-3 would affect CTL activity of G3KO CD8⁺ T cells. We found that G3KO CD8⁺ T cells were capable of expressing a near normal level of FasL, granzyme B, and perforin when fully activated with a high dose of anti-CD3 (10 μg/ml) in the presence of IL-2 (Fig. 7A). In addition, they were as efficient as FF cells in killing MHC-mismatched targets in vitro (Supplemental Fig. 4A). We then examined their Ag-specific CTL activity in an in vivo tumor model. We activated FF/OT-1 and G3KO/OT-1 cells in vitro with ova Ag at a dose that was high enough to activate both FF and KO cells comparably based on the value of FSC. The level of CD25 was, however, still slightly lower in G3KO cells even under the optimal condition (Supplemental Fig. 4B). We then intradermally injected B16-OVA melanoma cells, which were engineered to express the OVA Ag, into three groups of WT C57BL/6 mice. The first group of mice received no additional cell transfer; the second group received preactivated FF/OT-1 cells 4 d later; and the third group received preactivated G3KO/OT-1 cells. In the absence of OT-1 cells, B16-OVA cells expanded quickly and formed tumors of ∼1.6 cm in diameter by day 16 (Fig. 7B, left panel). FF/OT-1 cells were able to limit the growth of B16-OVA cells, and the average tumor size in mice of the second group was 0.7 cm in diameter. However, G3KO/OT-1 cells were inefficient at killing B16-OVA cells. The average tumor size in the third group was 1.4 cm in diameter. There were significantly fewer G3KO/OT-1 cells, comparing to FF/OT-1, within tumors (Fig. 7B, right panel). Tumor-infiltrating G3KO/OT-1 cells were smaller in size, judged by FSC-H, than infiltrating FF/OT-1 cells and still expressed a lower level of CD25, suggesting that they were less activated (Fig. 7C). Taken together, our data indicate that G3KO CD8⁺ T cells, despite their nearly normal CTL activity in vitro, were unable to maintain their activation status and were less efficient in killing Ag-specific target cells in vivo.

GATA-3 plays an important role in several steps during the differentiation of the T linage and in many subsets of T cells, including Th2 (19), NKT (26, 27), and most recently regulatory T (Treg) cells (28, 29). Our data further expand the role of GATA-3 in T lineage and demonstrate that GATA-3 regulates many biological aspects of CD8⁺ T cells.

Single-positive thymocytes downregulate CD24 prior to emigrating out of thymus. The transcriptional regulation of this final maturation step, however, is poorly understood. A recent article showed that deficiency of NKAP, a putative transcriptional repressor, led to impaired downregulation of CD24 in mature SP thymocytes (30). One possible scenario is that GATA-3 and NKAP act sequentially in the same molecular pathway that suppresses the expression of CD24. We found that the transcript level of NKAP was normal in G3KO SP thymocytes and naive CD8⁺ T cells (data not shown). In addition, other features of NKAP-deficient T cells are quite different from those of G3KO CD8⁺ T cells. For example, G3KO peripheral CD8⁺ T cells display a surface phenotype (CD24^loQa2^hi) indicating fully maturation in thymus (data not shown). This is in sharp contrast to NKAP-deficient naive peripheral T cells, which are CD24^hiQa2^lo. Therefore, GATA-3 probably regulates the final maturation of SP thymocytes by a mechanism different from that mediated by NKAP. A similar defect in downregulation of CD24 has also been observed in thymocytes defective in NK-κB signaling pathway and in OT-1/gld CD8SP thymocytes (31, 32). Further studies will be needed to elucidate how GATA-3 controls the downregulation of CD24 in SP thymocytes and whether there is any link between GATA-3, NF-κB, and/or Fas/FasL signaling pathways. Others and we have previously shown that deficiency of GATA-3 had only a modest impact on MHC class II–mediated positive selection (8, 9). We were surprised to see the profound defect in the positive selection of G3KO/OT-I thymocytes (Fig. 1). It remains to be determined whether this defect in positive selection can also be seen with other MHC class I–restricted TCR transgenes.

Our data revealed previously unknown role of GATA-3 in regulating CTL activity. Although adequately activated G3KO CD8⁺ T cells were competent in killing MHC-mismatched targets in vitro (Supplemental Fig. 4A), they were functionally impaired in killing Ag-expressing tumor cells in vivo. One possible explanation for the impaired in vivo killing is that G3KO CD8⁺ T cells have abnormal homing ability. There were indeed fewer tumor-infiltrating G3KO/OT-1 CD8⁺ T cells than FF/OT-1 CD8⁺ T cells. We cannot completely rule out this possibility. However, we showed in Fig. 3B and 3C that adoptively transferred G3KO CD8⁺ T cells or CD8SP thymocytes did home to secondary lymphoid organs appropriately. Given the observation that tumor-infiltrating G3KO/OT-1 CD8⁺ T cells were relatively smaller and expressed a lower level of CD25 than FF/OT-1 cells, we speculate that activated G3KO/OT-1 CD8⁺ T cells could not maintain their status of activation and accordingly were unable to efficiently expand, survive, and kill Ag-bearing tumor cells in vivo. It remains unclear whether the intrinsic CTL function of G3KO CD8⁺ T cells is intact or impaired in vivo.

Unlike T cells that are deficient in known signaling molecules, the initial signaling cascade after Ag encounter proceeded nearly uneventfully in G3KO T cells but the activation signals could not be sustained in the absence of GATA-3. Very little is known regarding how activation signals, once initiated, are maintained at the level of transcription. GATA-3, to the best of our knowledge, is the first transcription factor that is demonstrated to be required for the maintenance/progression, but not the initiation, of activation signals. However, its mechanism of action is still unclear. Our data indicate that lack of IL-2/STAT5 signaling is not the cause. Instead, GATA-3 is required for maintaining the protein level of a subset of signaling molecules, such as Lck, Zap70, Lat, and Sos, after the initiation of TCR signals. Although GATA-3 may directly control the expression of these signaling molecules, we seriously doubt this explanation for two reasons. First, the dependence on GATA-3 for maintaining the expression of these signaling molecules does not correlate with in vivo binding of GATA-3 to corresponding gene loci. Second, the level of these proteins is normal in naive cells prior to stimulation. Therefore, their dependence on GATA-3 is temporally regulated, and GATA-3 more likely indirectly controls the expression of these genes. More cells undergoing early apoptosis were detected in G3KO population before or immediately after stimulation, and the impaired activation we saw could be a consequence of enhanced apoptosis. In agreement with this scenario, several genes that are known to regulate the survival of T cells, including Mcl1, Fas, FasL, and Casp3, contain GATA-3 binding sites according to the ChIP-seq database (25). Although this scenario cannot be ruled out, we found that even the remaining healthy (7AAD⁻Annexin-V⁻)-looking G3KO cells were still poorly activated. Therefore, we speculate that the inability to maintain activation signals could be independent from or even the cause of enhanced apoptosis. More studies will be needed to test these hypotheses.

Whether GATA-3 is also required for the maintenance/progression of activation signals in other subsets of T cells or at different developmental stages of T lineage is not known. It was recently shown that G3KO Treg cells were functionally impaired. These cells could not maintain a high level of Foxp3 and were prone to reverting to effector cells. One striking feature of G3KO Treg cells was that they had impaired homeostasis and were unable to compete with control cells in mixed bone marrow chimeric mice. This defect in homeostatic expansion probably contributed to their inability to protect RAGKO mice from developing colitis induced by naive WT Th cells (28, 29). This defect is reminiscent of the impaired homeostatic expansion of G3KO CD8⁺ T cells we observed in RAGKO mice and may be also due to a disruption in the maintenance/progression of activation signals. The impaired signal maintenance/progression in the absence of GATA-3 may also explain why G3KO mice have virtually no CD4SP thymocytes despite normal initiation of TCR signaling in DP cells and a near normal number of polyclonal CD4^hiCD8^lo postselected cells expressing a near normal level of CD69 (8, 9), a marker of positive selection. It has been proposed that specification of CD4 lineage requires extended duration of TCR signals during MHC-mediated positive selection. It is logical to assume that impaired signal maintenance/progression caused by GATA-3 deficiency could lead to premature termination of TCR signals, thereby selectively interfering with the development of CD4SP cells.

We observed G3KO mice paradoxically developing age-dependent lymphadenopathy, despite a profound and cell-intrinsic defect in activation. One obvious and plausible explanation is lack of Treg cells. Unlike scurfy or Foxp3-deficient mice, the lymphadenopathy observed in G3KO mice is short-lived. It is possible that there still are few Treg cells present in young G3KO mice. This small Treg population then expands as G3KO mice age, thereby quenching the lymphadenopathy and abnormal expansion of CD8⁺ T cells in older mice. However, mice deficient in CD4, MHC class II, or Th-POK have virtually no CD4⁺ T cells but are not reported to have age-dependent lymphadenopathy or abnormal expansion of CD8⁺ T and B cells. Thus, simply lacking CD4⁺ T cells, including Treg cells, is probably not the cause of the lymphadenopathy seen in G3KO mice. This phenotype is even more striking in G3KO/OT-1 mice and G3KO mice never develop autoimmune features even up to 1 y of age, arguing against an autoantigen-driven process. The deletion of GATA-3 is mediated by CD4cre in our G3KO mice. The expression of CD4 is also detected in non-Th cells, including CD11c⁺ dendritic cells and lymphoid tissue inducer cells (33–35). One intriguing scenario is that some of these cells may also express GATA-3 and that the deletion of GATA-3 in these cells may drive the age-dependent expansion of CD8⁺ T cells in G3KO mice. These possible scenarios are being investigated.

In conclusion, absence of GATA-3 in CD8⁺ T cells caused an unexpected defect in cell-intrinsic proliferation and defect in vivo killing. A lack of sustained signals via the TCR may be responsible and indicates potential commonalities in the roles that GATA-3 plays in multiple sublineages of T cells.

The authors have no financial conflicts of interest.