Requirement for T-bet in the aberrant differentiation of unhelped memory CD8⁺ T cells

During the acute phase of an infection, repression of IL-7Rα marks pathogen-specific CD8⁺ T cells destined for elimination, whereas the cells that will give rise to self-renewing memory CD8⁺ T cells seem to be contained within the IL-7Rα–expressing subset (7, 8). When we examined the lymphocytic choriomeningitis virus (LCMV)–specific CD8⁺ T cell response in T-bet–deficient mice, we observed defective repression of IL-7Rα at day 8 after infection (Fig. 1 A). Because this finding suggested that T-bet may function as a repressor of IL-7Rα, we tested whether T-bet was sufficient to repress IL-7Rα in T cells. Retroviral-mediated expression of T-bet in CD8⁺ or CD4⁺ T cells stimulated in vitro resulted in repression of IL-7Rα mRNA (Fig. 1 B). In addition, ectopic expression of T-bet in developing Th2 cells in which T-bet is not normally expressed resulted in repression of IL-7Rα surface expression (Fig. 1 C).

The identification of T-bet as a repressor of IL-7Rα prompted us to test whether T-bet expression is associated with the IL-7Rα^lo subset of effector CD8⁺ T cells. We sorted LCMV-specific P14 CD8⁺ T cells 8 d after infection on the basis of IL-7Rα expression and examined T-bet mRNA. T-bet was enriched in the IL-7Rα^lo subset of effector CD8⁺ T cells, whereas the IL-7Rα^hi subset had reduced expression of T-bet mRNA (Fig. 1 D) and protein (not depicted). A similar pattern of gene expression was observed in IL-7Rα^hi and IL-7Rα^lo cells from the endogenous LCMV-specific CD8⁺ T cell response (not depicted).

In addition to elevated expression of T-bet, the IL-7Rα^lo subset of effector CD8⁺ T cells exhibited substantial enrichment for KLRG1 mRNA (Fig. 1 D) and protein (not depicted). KLRG1 is an NK-like inhibitory receptor that marks replicative senescence in CD8⁺ T cells (7, 9) and whose expression in NK cells is dependent on T-bet (10). In contrast, IL-7Rα^lo effectors had reduced expression of CCR7 mRNA (Fig. 1 D), which is a chemokine receptor that represents a defining feature of central–memory CD8⁺ T cells (2, 3). In addition, we discovered that the IL-7Rα^lo subset exhibited elevated expression of Blimp-1 mRNA (Fig. 1 D), a transcriptional repressor that promotes terminal differentiation of plasma cells (11) and that may also function in T lymphocytes (12, 13). Thus, enhanced expression of T-bet appears to selectively mark IL-7Rα^lo effector CD8⁺ T cells, which exhibit features of terminal differentiation and which have previously been shown to fail to give rise to self-renewing memory cells (7, 8).

Cell transfer experiments have suggested that the IL-7Rα^lo subset of effector CD8⁺ T cells preferentially gives rise to effector–memory cells (7). Consistent with this potential lineage relationship, we found that T-bet was enriched in effector–memory CD8⁺ T cells (Fig. 1, E and F). The skewing of T-bet expression in effector– versus central–memory cells was also recently reported in human memory CD8⁺ T cells (14). Given the variety of haploinsufficient phenotypes resulting from hemizygous deletion of T-bet, it is possible that the differences in the amount of T-bet between IL-7Rα^lo versus IL-7Rα^hi effectors and central– versus effector–memory CD8⁺ T cells could be functionally relevant (15–19). Expression of Blimp-1 paralleled that of T-bet, with elevated expression in the effector–memory CD8⁺ T cell subset (Fig. 1, E and F). Eomes expression, in contrast, did not vary substantially between IL-7Rα^hi and IL-7Rα^lo (Fig. 1 D) or central– and effector–memory subsets (Fig. 1, E and F).

The expression pattern of T-bet and its role in repressing IL-7Rα suggested that T-bet might negatively regulate the development of central–memory CD8⁺ T cells. We found that loss of T-bet, in addition to causing derepression of IL-7Rα, resulted in effector CD8⁺ T cells that acquired several characteristics of central–memory cells, including high expression of CD27, low expression of KLRG1, and robust IL-2 production (Fig. 2, A and C) (16, 20). 30 d after infection, LCMV-specific CD8⁺ T cells from wild-type mice exhibited substantial heterogeneity (Fig. 2 B), which is consistent with the presence of both central– and effector–memory CD8⁺ T cells (2, 3, 5). In contrast, T-bet–deficient mice exhibited a predominance of central–memory CD8⁺ T cells (Fig. 2, B and C). T-bet–deficient memory CD8⁺ T cells also exhibited a gene expression pattern similar to central–memory cells, with increased CCR7, reduced Blimp-1, and elevated Eomes mRNA (Fig. 2, D and E).

It was previously suggested that T-bet functions as a positive regulator of memory CD8⁺ T cell development because T-bet–deficient mice were found to have decreased numbers of LCMV-specific memory CD8⁺ T cells in the blood and spleen (16). In the memory phase of the response (30 and 60 d after infection), we also found that Tbx21^−/− mice had fewer LCMV-specific CD8⁺ T cells in the blood and spleen (Fig. 2 F and Fig. S1). Despite the deficiency in the blood and spleen, Tbx21^−/− mice were found to have increased numbers of memory cells in the lymph nodes. This might be partially attributable to the elevated expression of CCR7 in Tbx21^−/− CD8⁺ T cells (Fig. 2 D). To ensure that the phenotype of T-bet deficiency was CD8⁺ T cell–intrinsic, we used a transfer system in which the behavior of both wild-type and Tbx21^−/− P14 CD8⁺ T cells could be studied within the same wild-type host (Fig. S2). This system recapitulated the phenotypic, functional, and anatomic characteristics observed in the endogenous CD8⁺ T cell response of Tbx21^−/− mice. Together, these results suggest that T-bet inhibits the formation of lymph node–homing, central–memory CD8⁺ T cells and positively regulates the development of effector–memory CD8⁺ T cells.

Two features of central–memory CD8⁺ T cells are their robust capacity for secondary proliferation and an ability to confer heightened resistance to reinfection. Therefore, we assessed these characteristics in Tbx21^−/− memory CD8⁺ T cells by transferring equal numbers of GP33-specific wild-type or Tbx21^−/− memory CD8⁺ T cells into naive wild-type recipients and challenging with Listeria monocytogenes expressing GP_33-41. T-bet–deficient memory CD8⁺ T cells exhibited severalfold greater reexpansion than wild-type cells upon rechallenge (Fig. 3 A) and were found to express IL-2 at a greater frequency (Fig. 3, A and B). Mice that received Tbx21^−/− memory CD8⁺ T cells also showed substantially reduced bacterial burdens in both spleen and liver compared with recipients of wild-type cells (Fig. 3 C). Thus, deletion of T-bet promotes the development of highly functional memory CD8⁺ T cells. Presently, we are unsure why the defect in effector–memory CD8⁺ T cells resulting from deficiency of T-bet appears more permanent in this study compared with prior findings (16), although the operational definitions of effector–memory differ (phenotypic/functional versus chemokine receptor) in the two studies.

A well-characterized model of defective memory CD8⁺ T cell development involves CD8⁺ T cell activation in the absence of CD4⁺ T cell help (21–25). We examined the properties of “unhelped” CD8⁺ T cells by transfer of P14 CD8⁺ T cells into Cd4^−/− recipients (Fig. 4) or recipients depleted of CD4⁺ T cells using monoclonal antibody injection (Fig. 5, Fig. S3, and not depicted).8 d after infection, there was no difference in the properties of P14 CD8⁺ T cells in wild-type compared with Cd4^−/− hosts (not depicted). Several weeks after infection, however, P14 CD8⁺ T cells from Cd4^−/− hosts began to manifest signs of aberrant memory differentiation, with impaired expression of CD62L, IL-7Rα, CD27, and IL-2, and elevated expression of KLRG1 (Fig. 4 A).

The memory cells arising in the absence of CD4⁺ T cell help appeared to be more effector–memory–like, thus contrasting with T-bet–deficient memory CD8⁺ T cells, which appear to be more central–memory–like. Using intranuclear staining of the T-bet protein, unhelped memory CD8⁺ T cells exhibited a detectable increase in expression of T-bet (Fig. 4 B), which specifically localized to the expanded effector–memory–like subset (CD62L^lo, IL-7Rα^lo, CD27^lo, KLRG1^hi; Fig. 4 C and Fig. S4). In addition to exhibiting elevated T-bet mRNA (Fig. 4 D), unhelped memory CD8⁺ T cells were found to express more Blimp-1 mRNA, but less CCR7 and Eomes mRNA (Fig. 4, D and E). Lack of CD4⁺ T cell help, thus, appears to impair central–memory and/or promote effector–memory CD8⁺ T cell development.

To discriminate whether elevated T-bet expression in unhelped memory CD8⁺ T cells plays a causal role in, or is simply a consequence of, their altered differentiation, we performed antibody depletion of CD4⁺ T cells from wild-type or Tbx21^−/− mice, followed by infection with LCMV. Again, we found that unhelped memory CD8⁺ T cells in wild-type mice exhibited a predominance of effector–memory–like cells (CD62L^lo, IL-7Rα^lo, CD27^lo_, and KLRG1^hi), with impaired production of IL-2 and CD40L (Fig. 5, A and B, and Fig. S3). Deletion of T-bet, however, prevented the phenotypic and functional defects associated with the lack of CD4⁺ T cell help (Fig. 5, Fig. S3, and Fig. S5), resulting in enhanced generation of central–memory CD8⁺ T cells, regardless of the availability of CD4⁺ T cell help. The rescue was apparent even when T-bet deficiency was confined to antigen-specific CD8⁺ T cells (Fig. S5). Moreover, delayed antagonism of T-bet after unhelped memory CD8⁺ T cell development was capable of reversing several aspects of their dysfunctional phenotype (Fig. S6). Thus, T-bet appears to be required for the aberrant differentiation of memory CD8⁺ T cells that occurs in the absence of CD4⁺ T cell help.

How T-bet and CD4⁺ T cells execute their opposing effects on the balance between central– versus effector–memory CD8⁺ T cells remains to be determined. In several different systems, transcription factors function as intrinsic modulators of self-renewal or terminal differentiation. In B lymphocytes, the transcriptional repressor Blimp-1 promotes terminal differentiation (plasma cell) at the expense of self-renewal (memory B cell) (11). In the Drosophila melanogaster neural stem cells, the homeodomain transcription factor Prospero functions as a similar switch by repressing genes required for self-renewal, such as stem cell and cell-cycle genes, while activating genes involved in terminal differentiation (26). Parallels from other biological systems may help elucidate whether T-bet similarly functions to promote terminal differentiation at the expense of self-renewal, which has been proposed to represent an essential distinction between effector– versus central–memory CD8⁺ T cells, respectively (2–7, 27).

The potential ability of T-bet to switch between two polar states of differentiation may explain numerous aspects of viral pathogenesis. Clonal deletion, for example, could be regarded as a “self-renewal deficiency state,” associated with extreme excess of T-bet expression resulting from prolonged antigenic stimulation. Likewise, the association of high precursor frequency with enhanced central–memory formation might be considered a “disorder of excessive self-renewal,” explained by diminished T-bet expression as a consequence of excessive competition for antigen (5). The potential role of T-bet in these processes is currently being investigated.

The excessive differentiation of effector–memory CD8⁺ T cells in the absence of CD4⁺ T cell help could involve direct effects, such as loss of soluble or membrane-bound signals typically delivered by CD4⁺ T cells to dendritic cells or CD8⁺ T cells, including IL-2, chemokines, or CD40L (28–30). Alternatively, loss of CD4⁺ T cell help could act more indirectly, for example, by altering kinetics of pathogen clearance and necessitating prolonged effector function (3). Although the abnormalities arising in the absence of CD4⁺ T cell help appear to be at least partially T-bet–dependent, it is anticipated there could be T-bet–independent derangements contributing to the aberrant differentiation of unhelped memory CD8⁺ T cells. Nonetheless, antagonism of T-bet might represent a novel therapeutic approach for preventing the dysfunctional programming of CD8⁺ T cells in patients with compromised CD4⁺ T cell function.

Mice were maintained and used in accordance with the University of Pennsylvania Institutional Animal Care and Use Guidelines. Wild-type C57BL/6, Tbx21^−/− (18), DO11.10 TCR transgenic, and P14 TCR transgenic mice were housed in specific pathogen-free conditions before use. Tbx21^−/− P14 TCR transgenic mice were generated by interbreeding Tbx21^−/− mice with P14 transgenic mice. For analysis of the endogenous CD8⁺ T cell response, C57BL/6 mice or Tbx21^−/− were infected with 2 × 10⁵ PFU of LCMV Armstrong. H-2D^bGP33 or H-2D^bNP396 tetramers were used to identify LCMV-specific CD8⁺ T cells. For P14 transfer experiments, ∼5 × 10⁴ CD8⁺ T cells from naive P14 TCR transgenic mice (Thy1.1/1.1) were transferred intravenously into C57BL/6 or Cd4^−/− recipients (Thy1.2/1.2). For P14 cotransfer experiments, ∼5 × 10⁴ wild-type P14 cells (Thy1.1/1.1) were mixed with ∼5 × 10⁴Tbx21^−/− P14 cells (Thy1.1/1.2) and transferred to wild-type C57BL/6 recipients (Thy1.2/1.2). 1 d after transfer, recipient mice were infected with LCMV. For CD4 depletion experiments, wild-type or Tbx21^−/− mice received intraperitoneal injection of GK1.5 antibody (0.2 mg in PBS) 1 d before and 1 d after infection with LCMV. GK1.5 treatment resulted in efficient depletion of CD4⁺ T cells for >1 wk (not depicted). For Listeria monocytogenes experiments, mice were challenged with 2.5 × 10⁵ CFU of L. monocytogenes with transgenic expression of the LCMV-derived peptide GP_33-41. 4 d after bacterial challenge, spleens and livers were homogenized in 1% Triton X-100 in PBS, followed by limiting dilution culture on blood heart infusion agar.

Surface staining, intracellular cytokine staining, and flow cytometry were performed as previously described (18). For LCMV-derived peptide stimulations, splenocytes or lymph node cells were stimulated for 6 h with 0.2 μg/ml GP_33-41 or NP_396-404 peptide in the presence of 1 μg/ml brefeldin A. Antibodies used for flow cytometry were purchased from BD Biosciences. Intranuclear T-bet staining (Santa Cruz Biotechnology) was performed by fixation with 4% paraformaldehyde in PBS, followed by permeabilization and staining in 0.1% Triton X-100 and 1% FBS in PBS. Splenocytes from DO11.10 or P14 TCR transgenic mice were stimulated and transduced as previously described (18).

RNA extraction, cDNA synthesis, and quantitative real-time RT-PCR were performed as previously described (18). Primer and probe sets used for HPRT and Eomes detection were previously described (18). Presynthesized Taqman Gene Expression Assays (Applied Biosystems) were used to amplify the following sequences (gene symbols and Applied Biosystems primer set numbers in parentheses): Blimp-1 (Prdm1; Mm00476128_m1), CCR7 (Ccr7; Mm01301785_m1), KLRG1 (Klrg1; Mm00516879_m1), and T-bet (Tbx21; Mm00450960_m1). ‘Test’ gene values are expressed relative to that of HPRT, with the lowest experimental value standardized at 1.

Fig. S1 shows that T-bet–deficient antigen-specific memory CD8⁺ T cells preferentially accumulate in lymph nodes. Fig. S2 shows that T-bet represses central–memory development in a CD8⁺ T cell–intrinsic manner. Fig. S3 shows that deletion of T-bet rescues the phenotype and function of unhelped memory CD8⁺ T cells. Fig. S4 shows that unhelped memory CD8⁺ T cells have elevated expression of T-bet. Fig. S5 shows that T-bet deficiency restricted to antigen-specific CD8⁺ T cells is capable of correcting abnormalities associated with unhelped memory cells. Fig. S6 shows that preexisting dysfunction in unhelped memory CD8⁺ T cells can be reversed by antagonism of T-bet. The online version of this article is available.

We are grateful to E. Allenspach, J. Chang, C. DeJong, D. Knoblock, S. Gordon, I. Kinjyo, A. O'Hara, V. Palanivel, H. Shin, and K. Vinup for assistance and discussion.

These studies were supported by National Institutes of Health grants (AI042370, AI061699, AI071309, and AI007532) and the Abramson Family Cancer Research Institute.

The authors have no conflicting financial interests.