TOX Is Required for Development of the CD4 T Cell Lineage Gene Program

Transition of a multipotent precursor cell to a fully differentiated cell requires induction of a gene program that both enables specialized cell functions and represses alternative cell fates. This is the case for a key cell fate bifurcation of developing T lymphocytes in the thymus, where immature CD4⁺CD8⁺ (double positive [DP]) thymocytes are induced by TCR binding to MHC class I or class II molecules to differentiate into either the CD8 or CD4 T cell lineage, respectively. Lineage-specific transcriptional regulators, including ThPOK and GATA3 for the CD4 lineage and Runx3 and MAZR for the CD8 lineage, play essential roles in this process (1–3).

Runx3 and its closely related family member Runx1 function to silence expression of the Cd4 gene in CD8 lineage cells (one aspect of alternative fate repression in this instance) and early CD4⁻CD8⁻ thymocytes, respectively (4). Despite these differential in vivo functional activities, Runx1 can partially compensate for loss of Runx3, because CD8 T cell development is most efficiently blocked if both Runx1 and Runx3 genes, or their shared protein partner Cbfβ, are disrupted (5, 6). Misdirection of MHC class I-specific cells to the CD4 lineage in mice deficient in both Runx1 and 3 is observed, also consistent with a role for Runx activity in alternative fate repression (5, 7). However, expression of a Runx3 transgene is not sufficient to redirect MHC class II-restricted cells into the CD8 lineage (8), suggesting either that additional factors may be needed to activate CD8 lineage specification and/or to inhibit CD4 commitment or that timing of expression of this factor is critical. Although expression of MAZR is not lineage-specific, it, like Runx3 (7), can bind the Thpok silencer, helping to enforce the CD8 cell fate in MHC class I-specific developing thymocytes (3). In addition to its CD4-silencing activity, Runx3 is known to positively regulate other aspects of the CD8 lineage program, including regulating expression of CD8 itself (4, 9), thus fulfilling roles as both an alternative fate repressor and a positive cell fate regulator. Interestingly, this dual activity correlates with Runx-mediated induced proximity of the Cd4 and Cd8 loci (10). Runx1, unlike Runx3, is also expressed in developing CD4 lineage thymocytes where it plays a role in IL-7R expression and cell survival (5).

For CD4 T lineage development, expression of ThPOK (encoded by the Zbtb7b locus, referred to as Thpok in this article) is key. ThPOK is a zinc-finger BTB-POZ domain containing transcriptional regulator that, when absent, leads to misdirection of MHC class II-specific cells into the CD8 lineage and when ectopically expressed can misdirect MHC class I-restricted cells into the CD4 lineage, demonstrating the pivotal role of this nuclear protein in the CD4/CD8 lineage decision (11–14). GATA3 is more highly expressed in CD4 lineage thymocytes than in CD8 lineage thymocytes and is similarly required for CD4, but not CD8, lineage development (15, 16). However, unlike ThPOK, the absence or overexpression of GATA3 leads to inefficient lineage misdirection of developing thymocytes (15, 16). The upregulation of GATA3 precedes that of ThPOK, and, in the absence of GATA3, ThPOK fails to be induced (17, 18). Given that GATA3 binds the Thpok locus, it may play a direct role in ThPOK regulation, although overexpression of GATA3 is not sufficient to induce ThPOK expression in DP thymocytes (19).

Commitment to the CD4 and CD8 lineages requires that ThPOK and Runx3 be ultimately expressed in a mutually exclusive fashion. This, at least in part, appears to be achieved by the cross-inhibitory activities of these proteins; each can repress expression of the other (6, 7, 20, 21). Moreover, ThPOK can antagonize the Cd4-silencing activity of Runx3 (22) and inhibits juxtaposition of the Cd4 and Cd8 loci (10). The question then arises as to whether ThPOK, like Runx3 for the CD8 lineage, is also a positive regulator of the CD4 T cell fate. In the absence of both ThPOK and Runx activity, the latter achieved by conditional knockout of Cbfb, thymocytes develop into the CD4 lineage (6). This argues against an obligatory role for ThPOK as a positive regulator of the CD4 lineage. As noted, GATA3 has been implicated as a ThPOK regulator. However, the fact that ThPOK cannot restore CD4 T cell development in the absence of GATA3 (17) suggests other functions for GATA3, including the possibility of a role in induction of CD4 lineage gene expression and/or cell survival. Overexpression and ectopic expression of transgene-encoded GATA3 induces many changes in gene expression in MHC class II-specific TCR transgenic (Tg) thymocytes, including downregulation of RAG genes and modulation of various molecules that might influence TCR-mediated signaling (15, 23). However, whether GATA3 is also a direct inducer of the CD4 T cell lineage gene program remains uncertain.

Unlike ThPOK and GATA3, the pattern of expression of the high mobility group box nuclear protein TOX is not CD4 lineage biased (24). Nevertheless, TOX is required for CD4 T cell development, mediating progression of thymocytes from the CD4^loCD8^lo (double dull [DD]) to CD4⁺CD8^lo stages (25). Despite the absence of CD4⁺CD8^lo transitional cells in mice that lack TOX, CD8 T cell development is only modestly inhibited (25). However, in contrast to loss of ThPOK, there is little, if any, misdirection of MHC class II-specific T cells into the CD8 lineage in the absence of TOX (25). In this study, we sought to investigate whether the impaired development of CD4 T cells in TOX-deficient cells was solely due to an inability to induce ThPOK or, alternatively, whether TOX might also be one aspect of the regulatory network required for establishment of CD4 lineage gene expression. To study this, we expressed a ThPOK transgene in TOX-deficient (Tox^−/−) thymocytes. Although expression of ThPOK was sufficient to efficiently inhibit the CD8 lineage cell fate, thymic induction of many CD4 T cell lineage genes, including Thpok, Cd4, Id2, and Foxo1, as well as downstream gene targets, remained impaired. These data implicated TOX as a key factor in establishment of the CD4 lineage gene program.

All mice were bred at the Cedars-Sinai Medical Center and kept under specific pathogen-free conditions. AND TCR-Tg (26), Tox^−/− (25), and ThPOK-Tg (13) mice were described previously. AND-Tg Tox^−/− mice were kept on an MHC^b/b genetic background and screened for the absence of mouse mammary tumor virus 3 and 13. All mouse studies were reviewed and approved by the Cedars-Sinai Medical Center Institutional Animal Care and Use Committee.

All Abs used for flow cytometry were purchased from eBioscience or BD Biosciences. Anti-TOX (24) and anti-Runx (27) Abs were described previously, and anti-FOXO1 Ab was purchased from Cell Signaling Technologies. GATA3 intracellular staining was performed as per the manufacturer’s recommendation (clone L50-823; BD Biosciences).

CD4-prestained thymocytes or total splenocytes were stimulated with PMA (0.015 or 50 ng/ml, respectively) and ionomycin (0.3 or 500 ng/ml, respectively) in vitro for 4 h, and expression of CD154 or CD69 was assessed by flow cytometry.

Five-micrometer frozen sections from thymus were stained as indicated (along with DAPI for visualization of nuclei) and analyzed on a Leica TCS SP spectral confocal microscope using CytoView software (Infinicyte).

Real-time quantitative RT-PCR analysis was performed using the standard curve method, where samples were normalized based on Gapdh expression (and for enriched splenic samples, population purity as well) and analyzed using SDS 2.1 software (Applied Biosystems). Each gene-expression determination was performed in at least duplicate, and at least two independent biological experiments were performed. Primers for real-time RT-PCR were purchased from QIAGEN, with the exception of the Thpok 3′untranslated region (UTR) pair: 5′-CCCTTCCGAGACCCCTCCTTCCTGGT-3′ and 5′-CAATCAGGCCACCCTGCCCTCCTCTT-3′.

The p value associated with a Student t test using a two-tailed distribution of equal variance is shown in some figures.

Although spleen cellularity is normal in Tox^−/− mice, the T cell developmental defect in these animals results in T cell lymphopenia, primarily because of the loss of CD4⁺ T cells (Fig. 1A) (25). Despite this, TCR⁺CD4⁺ T cells can be detected in the spleen of these mice with age (Fig. 1A). In addition, we noted the presence of splenic CD4⁺CD8^int T cells in Tox^−/− mice that were not present in spleens from wild-type animals (Fig. 1A). Gene-expression analysis in these subsets revealed that Tox^−/− CD4⁺ and CD4⁺CD8^int T cells expressed lower amounts of Thpok mRNA compared with wild-type CD4⁺ T cells (Fig. 1B). Like their wild-type counterparts, CD8⁺ Tox^−/− T cells do not express Thpok (Fig. 1B). We also analyzed expression of the integrin CD103 on peripheral T cells. CD103 is normally expressed on ∼30% of splenic CD8⁺ T cells, but few CD4⁺ T cells (Fig. 1C) (28, 29). However, in the absence of TOX, CD103 was expressed by CD8⁺ T cells, as well as by a similar proportion of CD4⁺ T cells; ∼40% of CD4⁺CD8^int T cells and ∼30% of CD4⁺CD8⁻ single-positive (CD4SP) T cells expressed CD103 in these mice (Fig. 1C). The induction of CD103 on CD8 lineage thymocytes is regulated by Runx3 (30). Because ThPOK was shown to negatively regulate Runx3, the aberrant expression of the CD8 marker CD103 on the surface of Tox^−/− CD4⁺ T cells may be due to the impaired expression of ThPOK (and see below).

To determine whether we could detect Thpok expression under conditions where there was extensive generation of CD4 lineage T cells, we crossed Tox^−/− mice with mice expressing the MHC class II-specific AND TCR (AND-Tg) (26). CD4SP thymocytes failed to develop in AND-Tg Tox^−/− mice (Fig. 1D). The fact that few CD4⁻8⁺ single positive (CD8SP) thymocytes develop in AND-Tg Tox^−/− mice is consistent with the lack of extensive lineage misdirection in Tox^−/− mice on an otherwise wild-type background (25). To verify that TOX deficiency did not inhibit the initiation of positive selection in these animals, we analyzed expression of TCR and CD5. Both TCR and CD5 are expressed at low levels on DP thymocytes and are upregulated during initiation of positive selection as a consequence of TCR signaling (Fig. 1E) (31). However, in AND-Tg thymocytes, CD5 is upregulated on most thymocytes, including both DP and CD4SP cells, as a result of global positive selection that results from uniform and early expression of the AND TCR (Fig. 1E). In AND-Tg Tox^−/− mice, the majority of cells also exhibited upregulated levels of CD5, although development of TCR^hi CD4SP cells was still impaired (Fig. 1E).

The differential effect of loss of GATA3 in development of the CD4 and CD8 T cell lineages is mirrored by the expression pattern of the protein, because GATA3 is more highly expressed in CD4SP thymocytes than in CD8SP thymocytes (15, 16). Gata3 gene expression was indistinguishable in total thymocytes from Tox^−/− mice compared with thymocytes from Tox^+/+ or Tox^+/− mice (Fig. 1F). Thus, as observed using CD5 as a marker, the majority of DP thymocytes in AND-Tg mice have received initial positive-selection signals independent of TOX expression. Therefore, Gata3 and Tox are independently regulated to a significant degree. Although upregulation of Tox is initiated by TCR-mediated calcineurin signaling during positive selection (32), the regulation of GATA3 may instead be dependent, at least in part, on MAPK (15, 33). However, despite induction of GATA3, AND-Tg Tox^−/− thymocytes did not express Thpok (Fig. 1G). Thus, ThPOK fails to be induced in AND-Tg mice in the absence of TOX, despite initiation of positive selection and upregulation of GATA3.

These data raised the possibility that the failure to upregulate ThPOK was critical for TOX-dependent CD4 T cell development. To test this, we bred Tg mice that express ThPOK (ThPOK-Tg) onto a Tox^−/− background. Analysis of surface expression of TCR and CD5 suggested that ThPOK-Tg Tox^−/− cells displayed a postpositive selection block identical to that observed in the thymus of Tox^−/− mice (Fig. 2A) (25). Thus, although initiation of positive selection, as assessed by the presence of a CD5^hiTCR^int population, was evident in both Tox^−/− and ThPOK-Tg Tox^−/− mice, CD5^hiTCR^hi thymocytes were markedly reduced in mice of either genotype.

However, subset analysis based on coreceptor expression presented a very different picture of the effect of enforced ThPOK expression. CD8SP thymocytes fail to develop in ThPOK-Tg mice as a result of misdirection of all positively selected thymocytes to the CD4 T cell lineage (13, 14). Similarly, CD4SP, but not CD8SP, thymocytes are present in ThPOK-Tg Tox^+/− mice (Fig. 2B, 2C). Although the frequency and number of DP thymocytes is not diminished in Tox^−/− mice, CD4SP thymocytes fail to develop, and there is expansion of the postselection DD subpopulation in these animals (Fig. 2B) (25). This is in contrast to TCR⁺ CD8SP thymocytes, which develop in the absence of TOX (Fig. 2B) (25). However, in Tox^−/− mice that also expressed the ThPOK transgene, TCR⁺ CD8SP cells were eliminated (Fig. 2B, 2C). In addition, although ThPOK-Tg Tox^−/− mice lacked CD4SP thymocytes like their Tox^−/− counterparts, a subset of cells that downregulated CD8, but failed to fully upregulate CD4 (CD4^loSP), was present instead (Fig. 2B, 2C). Despite their unusual phenotype, CD4^loSP thymocytes from ThPOK-Tg Tox^−/− mice expressed CD5 and GATA3 equivalently to that observed in CD4SP thymocytes from ThPOK-Tg Tox^+/− mice (Fig. 2D, 2E), although expression of TCR and CD69 was lower and similar to that of DD thymocytes (Fig. 2D, data not shown).

Enforced expression of ThPOK resulted in the return to a normal lineage-restricted expression pattern of CD103 in Tox^−/− mice, because ThPOK-Tg Tox^+/− CD4SP and Tox^−/− CD4^loSP cells did not express CD103 (Fig. 3A). Tox^−/− CD8SP and, to a significantly lesser degree, DD thymocytes express Runx3 protein (25). However, like their Tox^+/− counterparts, DP and CD4^loSP ThPOK-Tg Tox^−/− thymocytes do not express Runx3 (Fig. 3B).

Enforced expression of ThPOK resulted in normal splenic T cell numbers (Fig. 3C, 3E) but loss of the CD8 T cell subset as a result of shunting of these cells into the CD4 T cell lineage during thymic development (Fig. 3D, 3F). In contrast, ThPOK-Tg Tox^−/− mice displayed T cell lymphopenia similar to that seen in Tox^−/− mice (Fig. 3D, 3E), although splenic CD8⁺ T cells were replaced by CD4⁺ T cells (Fig. 3D, 3F). As noted above, an abnormal subset of CD4⁺CD8^int T cells was found in Tox^−/− mice (Figs. 1A, 3D, 3G). These cells were diminished when ThPOK was also expressed (Fig. 3D, 3G). As expected, expression of Prf1 was specific to the CD8 T cell lineage in wild-type mice, and CD4⁺ T cells from ThPOK-Tg Tox^+/− mice also failed to express this CD8 lineage marker (Fig. 3H). Expression of Prf1 was also repressed in CD4⁺ T cells from ThPOK-Tg Tox^−/− mice (Fig. 3H). Together, these data demonstrated that repression of the CD8 lineage is ThPOK dependent but TOX independent.

To determine whether the CD4 lineage program was induced in ThPOK-Tg Tox^−/− CD4⁺ T cells, we surveyed some hallmarks of CD4 lineage commitment, including expression of the coreceptor itself. The DD phenotype is a reflection of coreceptor gene expression. Thus, the Cd4 gene is poorly expressed by Tox^−/− DD thymocytes compared with DP thymocytes (Fig. 4A). However, despite expression of ThPOK, ThPOK-Tg Tox^−/− CD4^loSP cells did not reactivate transcription of the Cd4 gene to normal levels (Fig. 4B). This is consistent with the TOX dependence of the DD to CD4⁺8^lo transition (25), a developmental progression that normally precedes ThPOK expression (13, 14).

One defining marker of the CD4 lineage is expression of ThPOK itself. To analyze this in ThPOK-Tg mice, we used a primer pair specific for the 3′UTR of Thpok, a region of the gene not included in the transgene. CD4SP thymocytes and CD4⁺ T cells in both Tox^+/− and ThPOK-Tg mice expressed endogenous Thpok (Fig. 4C, 4D). Therefore, expression of transgene-encoded ThPOK in the latter animals does not inhibit endogenous Thpok locus activation during CD4 lineage commitment. Indeed, expression of ThPOK may inhibit its own silencing in a positive-feedback loop (21). However, in sharp contrast, ThPOK-Tg Tox^−/− CD4^loSP thymocytes and CD4⁺ T cells poorly expressed endogenous Thpok, despite expression of transgene-encoded ThPOK in the cell (Fig. 4C, 4D).

Positive selection in the thymus requires a decrease in E protein transcriptional activity. This is regulated by upregulation of Id (inhibitor of DNA binding) proteins that lack DNA-binding capability and form nonfunctional dimers with E proteins, resulting in dominant negative inhibition of E protein activity. Loss of Id3 partially inhibits development of both CD4SP and CD8SP thymocytes (34), whereas direct loss of E protein expression, via dual conditional deficiency in HEB and E2A, promotes some changes in gene regulation shared with positive selection, even in the absence of TCR signaling (35). Like Id3, Id2 is upregulated during positive selection (data not shown), although positive selection is not inhibited in Id2-deficient mice (36, 37). To address whether loss of TOX altered the normal induction of Id proteins, we compared CD4SP and CD4^loSP thymocytes from ThPOK-Tg Tox^+/− and ThPOK-Tg Tox^−/− mice, respectively. In ThPOK-Tg Tox^−/− CD4^loSP thymocytes, expression of Id2 was significantly decreased in the absence of TOX (Fig. 4E), whereas no significant change in Id3 was detected (Fig. 4F).

A reduction in Id2 expression would be predicted to result in failure to fully downregulate E protein activity. As a consequence, one might expect loss of expression of genes that are repressed by E proteins and induced by the decrease in E protein activity during positive selection. To determine whether this was the case, we took advantage of gene-expression data from analysis of doubly deficient E2A and HEB mice (35). We noted that expression of the tescalcin gene (Tesc), encoding an EF-hand calcium-binding protein implicated as a regulator of plasma membrane-type Na⁺/H⁺ exchangers (38), cell differentiation programs (39), and possibly calcineurin activity (40), was highly upregulated in DP thymocytes upon loss of E protein activity. Moreover, we found that Tesc was highly upregulated at the CD4⁺8^lo transitional stage of positive selection (data not shown). Expression of Tesc was reduced in CD4^loSP thymocytes from ThPOK-Tg Tox^−/− mice compared with CD4SP thymocytes from ThPOK-Tg Tox^+/− mice (Fig 4G). Together, these results are consistent with a failure to fully repress E protein activity.

To extend these results to another gene target for which there was better functional characterization in the context of CD4 T cells, we analyzed expression of Foxo1, also upregulated upon loss of E proteins in DP thymocytes (35). FOXO1 is a transcriptional regulator that is induced during positive selection (41) and plays a key role in regulating cell migration and maintaining quiescence of mature T cells (42, 43). FOXO1 was significantly decreased at the level of both mRNA (Fig. 5A) and protein (Fig. 5B, 5C) in ThPOK-Tg Tox^−/− CD4^loSP thymocytes in comparison with ThPOK-Tg Tox^+/− CD4SP thymocytes. One target of FOXO1 is Klf2 (44). Consistent with reduction in FOXO1, Klf2 failed to be fully expressed in the absence of TOX when these same cell populations were compared (Fig. 5D). FOXO1 and KLF2 have many gene targets, including S1pr1, Il7r, Ccr7, and Sell (encoding l-selectin). Indeed, S1pr1 was significantly decreased in ThPOK-Tg Tox^−/− CD4^loSP thymocytes in comparison with ThPOK-Tg Tox^+/− CD4SP thymocytes (Fig. 5E). In addition, surface expression of CD127 (IL-7Rα), CCR7, and CD62L was decreased in the absence of TOX (Fig. 5F).

ThPOK-Tg Tox^−/− CD4^loSP thymocytes were also smaller than ThPOK-Tg Tox^+/− CD4SP thymocytes (Fig. 5G) and more similar in size to DD thymocytes (data not shown), suggesting a differentiation or maturation defect.

We also investigated CD4 lineage functional maturation by assessing CD40L (CD154) induction. Although activated ThPOK-Tg CD4SP thymocytes upregulated CD154 in response to PMA and ionomycin, their CD4^loSP Tox^−/− counterparts did not (Fig. 5H). Splenic CD4⁺ T cells in ThPOK-Tg Tox^−/− mice also had a defect in induction of CD154 upon activation (Fig. 5I, 5K). However, this was not due to loss of a downstream-signaling response, because the MAPK-dependent activation marker CD69 was upregulated upon activation, even in the absence of TOX (Fig. 5J, 5L). A higher frequency of ThPOK-Tg Tox^−/− CD4⁺ T cells was CD44^hi (Fig. 5N) and expressed CD69 (Fig. 5J, 5L) (also CD44^hi, data not shown) in the absence of stimulation compared with ThPOK-Tg Tox^+/− CD4⁺ T cells (Fig. 5J, 5L). Furthermore, there was a reduction in Foxo1 expression (Fig. 5M) compared with CD4⁺ T cells from ThPOK-Tg Tox^+/− mice, mirroring effects in the thymus. Together, these markers indicated that a significant proportion of CD4 T cells in ThPOK-Tg Tox^−/− mice were in an activated state. Upregulated expression of both CD44 and CD69 on naive CD4⁺ T cells was also reported in mice lacking FOXO1 (42). Thus, the reduction in FOXO1 in the absence of TOX is likely causative of the activated phenotype.

The decrease in expression of FOXO1 gene targets included factors responsible for thymocyte survival and migration. We previously demonstrated that loss of TOX is not associated with increased apoptosis in the thymus (25). Nevertheless, because bcl2 is a downstream gene target of IL-7R signaling in the thymus (45), and there was diminished expression of IL-7R in the absence of TOX, we asked whether expression of a bcl2 transgene might overcome the loss of TOX. Expression of BCL2 did not rescue the CD4 T cell developmental defect of Tox^−/− mice, although it did result in increased frequencies of DD and CD8SP subsets (Fig. 6A) (25). Consistent with the latter thymic phenotype, expression of BCL2 rescued the splenic T cell lymphopenia of Tox^−/− mice by increasing the frequency of CD8⁺ T cells (Fig. 6B). Although expression of CCR7, which functions in migration of single-positive (SP) thymocytes from the cortex into the medulla, and S1PR1, required for thymic egress, were impaired in ThPOK-Tg Tox^−/− mice, CD4⁺ T cells were still able to populate the spleen, although in reduced numbers (Fig. 5D, 5F). Impaired, but not absent, CD4 T cell maturation in these animals was also reflected by small medullary regions that nonetheless contained CD4SP thymocytes (Fig. 7).

In sum, despite CD8 lineage repression and the development of CD4⁺ T cells in ThPOK-Tg Tox^−/− mice, induction of the full CD4⁺ T lineage gene program, including those encoding proteins involved in cell differentiation, maturation, and function, failed to be established in the absence of TOX.

We previously found that CD4 T cell development is severely impaired in the absence of TOX (25), a nuclear protein predicted to be a structure-dependent but sequence-independent DNA-binding factor (46). The minor CD4 T cell population that escapes the thymus in the absence of TOX had lower than normal expression of Thpok, supporting a role for TOX in regulation of ThPOK expression. In addition, a population of splenic CD4⁺8^int T cells was present in Tox^−/− mice, and the normal lineage bias of CD103 expression was eliminated. CD4⁺8^int cells have also been reported in mice that express a hypomorphic allele of Thpok (6, 21), consistent with presence of these cells in Tox^−/− mice due to misregulation of ThPOK during thymic development. Indeed, CD8 repression in peripheral CD4⁺ T cells was demonstrated to require continued expression of ThPOK (6, 47).

Whether regulation of ThPOK by TOX is a direct or indirect effect remains to be determined. In terms of the latter, it is possible that TOX influences TCR signaling, likely needed for ThPOK induction (14). Although we have not observed any gross perturbations in TCR-mediated signaling in Tox^−/− thymocytes (25), reduced expression of CD4 and the TCR could be a factor. Alternatively, TOX could act independently on the ThPOK locus or render the locus more accessible to GATA3 (17) and/or other factors. Indeed, the fact that TOX is also expressed in cells developing to the CD8 lineage indicates that cell context and additional inputs likely modify the effects of TOX expression. Enforced expression of ThPOK eliminates expression of CD8, CD103, and Prf1 by Tox^−/− CD4⁺ T cells. ThPOK also repressed expression of Runx3, even in the absence of TOX, likely preventing CD103 expression (30). ThPOK can inhibit its own silencing, possibly by inhibition of Runx3, thus initiating a positive-feedback loop (21). Nevertheless, ThPOK and the lack of Runx3 expression were not sufficient to fully activate the Thpok locus in the absence of TOX. We concluded from these data that TOX plays a role in alternative fate repression solely as a modulator of ThPOK induction.

CD4⁺ T cells can develop, even in the absence of ThPOK, as long as the activity of Runx factors is also eliminated (6). This observation suggested that additional factors likely play a positive role in development of a CD4 lineage program. Our data suggested that TOX is one such factor. CD4⁺ T cells replaced CD8⁺ T cells in Tox^−/− mice when ThPOK was expressed. Despite this, CD4 lineage development in the thymus was compromised in these animals, including failure to fully re-express CD4 and poor induction of CD154 on SP thymocytes and CD4 T cells. The former observation is consistent with TOX dependence (25) but ThPOK independence (6, 12) of the DD to CD4⁺CD8^lo transition. Thus, one aspect of the requirement for TOX in mediating the development of CD4⁺CD8^lo cells may be regulation of the Cd4 locus itself.

Despite expression of CD4 on both DP thymocytes and CD4SP thymocytes, there are significant differences between these immature and mature cells in terms of accessibility, occupancy, and histone modification of the promoter, enhancer, and silencer elements (48–50). Although the DD subpopulation, in particular, has not been studied in this context, there is a report of silencer activity in these cells as assessed using marker Tg mice (48). Runx3 protein can be detected at low levels in DD thymocytes (25). In contrast, CD8 lineage T cells express high levels of Runx3 coincident with a switch to the distal promoter (5). It is possible that TOX plays a role in relief of this silencing, leading to full re-expression of CD4 and acquisition of the CD4⁺CD8^lo phenotype. If so, this effect would have to be cell context dependent, because expression of TOX in DP thymocytes can lead to Cd4 silencing (32). Alternatively, TOX may influence other regulatory elements, including that of a yet-to-be-identified Cd4 enhancer of mature T cells (50). In addition, modulation of E protein activity (see below), implicated in Cd4 gene regulation (51), might play a role in misregulation of CD4 in the absence of TOX.

The basic helix–loop–helix E protein transcriptional regulators E2A/E47 and HEB play multiple roles in the thymus (reviewed in Ref. 52). In terms of the DP to SP transition, inhibition of E protein activity mediated by TCR induction of Id3 plays a nonredundant role, particularly for CD4SP development (34). The normal expression of Id3, which is rapidly induced as a consequence of TCR activation of the MAPK pathway during positive selection (53), in Tox^−/− mice is consistent with unperturbed initiation of positive selection in the absence of TOX (25). In contrast, Id2 expression peaked at the later TOX-dependent CD4⁺8^lo transitional stage (data not shown), and expression of Id2 was diminished in ThPOK-Tg Tox^−/− CD4SP thymocytes.

A functional role for Id2 in the thymus has not been defined, but one might predict that upregulation would further diminish E protein activity. That the DP to SP transition is inhibited in Id3-deficient, but not Id2-deficient, mice may reflect the differences in kinetics of expression of these proteins. Because Id2 is upregulated after induction of Id3 during positive selection (data not shown), it would not be expected to compensate for loss of Id3, whereas Id3 may have compensatory activity when Id2 is absent. The finding of reduced Id2 in ThPOK-Tg Tox^−/− CD4SP thymocytes led us to ask whether there would be consequences for expression of genes that might normally be repressed by E protein activity. This was indeed the case because expression of both Tesc and Foxo1, genes previously shown to be upregulated in E2A/HEB doubly deficient thymocytes (35) and in the case of Foxo1, downregulated upon expression of E47 in a thymic lymphoma cell line (54), were poorly induced in the absence of TOX. The role of tescalcin in the thymus remains to be determined, although there is the potential to modulate cell signaling and/or cell differentiation (39, 40, 55). Better defined is the function of FOXO1 in CD4⁺ T cells (42). Expression of FOXO1 primary and downstream gene targets, including Klf2, S1pr1, Il7r, Ccr7, and Sell, was compromised in the absence of TOX. ThPOK-Tg Tox^−/− CD4SP thymocytes were still able to localize to medullary regions of the thymus, albeit inefficiently, and exit to the periphery, indicating that downregulation of these factors did not result in a complete block in chemotaxis or migration. Moreover, decreased FOXO1 expression in the absence of TOX was also associated with increased frequency of CD44^hi (and CD69^hi on a subset of these) CD4⁺ T cells in ThPOK-Tg Tox^−/− mice, as also observed in FOXO1-deficient mice (42).

It is intriguing that Tox^−/− mice, like Id2-deficient mice, have blocks in development of NK cells and LTi cells (56, 57). Tox^−/− bone marrow NK cells also have lower expression of Id2, similar to that reported in the thymus in this article (58). Nevertheless, ectopically expressed Id2 does not rescue development of Tox^−/− NK cells in vitro, pointing to additional roles for TOX in this process (58). Together, our data are consistent with TOX as an upstream modulator of Id2 expression, both in developing T cells and in other cell lineages. Expression of the E protein E47 in an E2A-deficient thymic lymphoma cell line leads to modest reduction in Tox expression (54), suggesting that E proteins could be an upstream repressor of TOX expression in some contexts. However, TOX and Id3 are induced by different signaling pathways (32, 53), and TOX expression precedes that of Id2 (data not shown). These data argued against the idea that a reduction in E protein expression at the initiation of positive selection is key to TOX induction. An alternate possibility is that there is a feed-forward regulatory loop where TOX-dependent upregulation of Id2 and ensuing downregulation of E protein activity enhance TOX expression. Indeed, TOX and Id2 expression both peak at the CD4⁺8^lo stage.

We propose a model whereby TOX is required for Cd4 re-expression at the DD to CD4⁺8^lo transition, as well as to create a permissive state for ThPOK induction, resulting in CD8 fate repression independent of the activity of TOX. In addition, suppression of E protein activity may occur in discrete stages during positive selection. TOX would play a role in Id2 upregulation and full E protein inhibition at the CD4⁺8^lo stage, necessary for induction of multiple aspects of the CD4 T cell lineage gene program, as well as to enhance TOX expression. The fact that TOX is not CD4 lineage specific points to the existence of other factors, potentially GATA3, that likely influence the biological function of TOX. Nevertheless, our data demonstrated that the “pulse” of TOX expression during positive selection is a key regulatory event in CD4 lineage differentiation.

We thank K. Wawrowsky for assistance with confocal microscopy, P. Lin and G. Hultin for cell sorting, and A. Seksenyan for critical reading of the manuscript. We are also grateful to M. Satake for the anti-Runx Ab and R. Bosselut for a CD154 thymocyte-staining protocol.

The authors have no financial conflicts of interest.