CD8αα-Mediated Survival and Differentiation of CD8 Memory T Cell Precursors

See allHide authors and affiliations

Science  23 Apr 2004:
Vol. 304, Issue 5670, pp. 590-593
DOI: 10.1126/science.1092316


Memory T cells are long-lived antigen-experienced T cells that are generally accepted to be direct descendants of proliferating primary effector cells. However, the factors that permit selective survival of these T cells are not well established. We show that homodimeric α chains of the CD8 molecule (CD8αα) are transiently induced on a selected subset of CD8αβ+ T cells upon antigenic stimulation. These CD8αα molecules promote the survival and differentiation of activated lymphocytes into memory CD8 T cells. Thus, memory precursors can be identified among primary effector cells and are selected for survival and differentiation by CD8αα.

The majority of T cells responding during a primary immune response subsequently undergo programmed cell death. However, a fraction of activated T cells survive and differentiate into long-lived memory T cells (1). What mechanisms mediate the selective survival of these cells? To address this question, we first must identify those effector T lymphocytes that will differentiate into memory cells.

The homotypic form of CD8 that uses the α chain of the molecule (CD8αα) appears to serve functions that are distinct from those of the T cell receptor (TCR) coreceptors CD4 and CD8αβ (24). Immature thymocytes can induce CD8αα upon strong TCR stimulation (57), and in mice (4, 811) and humans (12, 13), CD8αα is expressed on distinct T cell subsets that constitutively display a memory phenotype. In light of these characteristics, we hypothesized that CD8αα might have functional relevance in specifying T cell memory fate.

We recently showed that the thymic leukemia antigen TL, a nonclassical major histocompatibility complex (MHC) class I molecule, is a unique ligand for CD8αα, with TL tetramers binding specifically to CD8αα but not to CD8αβ (4). To determine whether mature T cells induce CD8αα after TCR stimulation with an antibody to CD3, we stained splenocytes with TL tetramer. Although no tetramer staining could be detected on resting splenocytes, the majority of CD8αβ+ T cells bound TL tetramer after polyclonal stimulation, and CD8αα disappeared by day 5 (Fig. 1A). CD8αα induction was also detected on a subset of CD8αβ+ OT-1 TCR transgenic splenocytes stimulated with specific peptide antigen (Fig. 1B). CD8αα expression was greatest on OT-1 splenocytes stimulated by antigen-presenting cells (APCs) that had been transfected with TL (APC/TL) (Fig. 1B); this result suggested that the interaction of CD8αα with TL might stabilize CD8αα surface expression.

Fig. 1.

CD8αα is transiently induced on activated CD8 splenocytes and provides survival. (A) C57BL/6 splenocytes cultured with antibody to CD3ϵ were analyzed for CD8αα expression using TL tetramers on activated (open) and resting (filled) CD8β+ and CD4+ T cells. (B) CD8+ OT-1 TCR splenocytes cultured with RMAS (APC) or RMAS-TL (APC/TL) with or without OVA peptide (OVAp) for 3 days were analyzed for CD8αα on gated OVA/H-2Kb tetramer+ cells. Data in (A) and (B) are from one of five independent experiments. (C) CD8+ OT-1 TCR splenocytes cultured with APC or APC/TL with or without OVAp for 72 hours. Apoptotic cells were detected with Bcl-xL and Annexin-V staining; blocking of Bcl-xL accumulation was observed in the presence of antibody 18/20 to TL (thin line). (D) After OVAp stimulation for 72 hours, CD8αα+ and CD8αα OT-1 TCR splenocytes were analyzed for IL-2Rα and IL-2/IL-15Rβ. Data are from one of four individual experiments. CD8 OT-1 TCR T cells were cultured with the indicated cytokines and OVAp-loaded irradiated wild-type spleen cells; the bar graphs show numbers of CD8αα+ cells on gated Vα2+ T cells (left) and numbers of CD8αα+ cells and total cells (right). Data are from one of three individual experiments. (E) Left panels: Total splenocytes, stimulated as above, were gated on CD11c+ or CD11b+ TCRβ cells and analyzed for TL expression using antibody HD168 to TL. Cells were resting (filled), stimulated (open, thick line), or in the presence of biotinylated antibody to rat isotype (open, thin line). Right panel: Annexin-V and TL tetramer staining of splenocytes stimulated with antibody to CD3ϵ for 72 hours. Data are from one of four independent experiments. Values in the horizontal bar represent percent positive cells for each staining.

Binding of classical MHC class I tetramers to DP thymocytes and activated T cells is known to be influenced by modified glycosylation on CD8αβ (14, 15). To exclude the possibility that this accounted for the observed TL tetramer staining on recently activated CD8αβ T cells, we used chimeric tetramers in which the CD8-binding α3 domain of TL was replaced by that of H2-Kb (TL/Kb). Under conditions where CD8αα was readily detected by TL tetramers, TL/Kb tetramers failed to stain activated OT-1 T cells (fig. S1). Additionally, TL tetramer staining of activated OT-1 cells could be blocked by antibodies to CD8α but not by antibodies to CD8β (fig. S1), demonstrating the specificity of the TL tetramer for CD8αα.

Previous studies have suggested that CD8αα might promote thymocyte survival (2, 7). We therefore hypothesized that CD8αα might also rescue mature activated CD8αβ T cells. Consistent with this idea, CD8αα+ splenic OT-1 transgenic T cells that had been activated by APC/TL retained high levels of the antiapoptotic factors Bcl-xL (Fig. 1C) and Bcl-2 (fig. S2) (16). The high levels of survival factors were dependent on TL expression by the APC and could be blocked using antibodies to TL (Fig. 1C). Accumulation of the antiapoptotic factors also correlated with enhanced lymphocyte survival, as measured by reduced uptake of Annexin V (Fig. 1C). Activated CD8αα+ OT-1 TCR splenocytes also expressed high levels of the shared interleukin-2 (IL-2)/IL-15 receptor β (Rβ) chain (also called CD122), but not the IL-2Rα chain (CD25) (Fig. 1D). This finding suggested that these cells expressed the IL-15 rather than the IL-2 receptor, and this was supported by the capacity of IL-15 to expand CD8αα+ OT-1 CD8+ splenocytes, as compared with an inhibition in the presence of IL-2 (Fig. 1D). Other cytokines, including IL-7, had no effect on the expansion of the CD8αα+ T cells (Fig. 1D). Costaining of polyclonal activated T cells with TL tetramers and Annexin V further demonstrated specific survival of the CD8αα+ population (Fig. 1E).

To relate this apparent survival advantage to the interaction of CD8αα with its ligand, we examined TL expression during splenocyte stimulation. Although TL was not detected on resting splenocytes, its expression was observed on activated CD3/CD11b/c+ dendritic cells and CD3/CD11b+/c monocytes (Fig. 1E), consistent with a potential CD8αα/TL-mediated survival process of the in vitro activated T cells.

To examine CD8αα induction on T lymphocytes in vivo, we infected mice with lymphocytic choriomeningitis virus (LCMV) and analyzed them at various times for expression of CD8αα on LCMV-reactive T lymphocytes (16). LCMV peptide–MHC class I tetramers were used to identify virus-specific T cells in conjunction with TL tetramers. Induction of CD8αα was observed on a subset of responding CD8αβ+ LCMV-specific cells by day 7 after infection (Fig. 2A, day 7). A proportion of the CD8αα+ responder cells did not bind the LCMV-specific tetramers, most likely because they represented activated T cells specific for other LCMV epitopes (Fig. 2A, day 7). The expression of CD8αα on the LCMV-specific CD8αβ T cells disappeared with virus clearance (Fig. 2A, days 14 and 40) (17). As was the case in vitro, activated LCMV-specific splenocytes that had induced surface CD8αα expression also expressed high levels of Bcl-xL (Fig. 2B), consistent with a selective survival advantage for these cells. In contrast to primary LCMV stimulation (Fig. 2A, day 7), nearly all the long-term surviving LCMV-specific CD8+ T cells reinduced CD8αα upon secondary stimulation (Fig. 2A, day 60). This is consistent with the possibility that the capacity to reinduce CD8αα was selected specifically during the primary response. The LCMV-specific primary activated CD8αα+ T cells also displayed enhanced expression of IL-7Rα (also called CD127) and IL-15Rβ (Fig. 2B), which are typically up-regulated on memory CD8 T cells (18, 19).

Fig. 2.

CD8αα is selectively induced on effector CD8 T cells in vivo. (A) Wild-type mice infected with LCMV were bled on the indicated days after infection (day 60 mice were rechallenged 3 days before analysis) and gated CD8+ cells were stained with GP276-286/Db and TL tetramers. (B) Responder cells isolated at day 7 after infection were analyzed for intracellular Bcl-xL and surface expression of IL-2/IL-15Rβ and IL-7Rα. Filled and open histograms denote CD8αα and CD8αα+ effector cells, respectively. Data are representative of two individual experiments in (A) and (B) with 30 mice in each experiment. Values shown in each quadrant of (A) are percentages of tetramer+ cells on gated CD8+ T cells; percentages of tetramer+ cells of total splenocytes are in parentheses.

Given the characteristics so far described for CD8αα+ activated lymphocytes, we reasoned that they might represent precursors of memory T cells. Naïve CD8+ splenocytes isolated from transgenic mice expressing the P14 TCR (specific for the LCMV GP33-41 epitope presented by H-2Db) were transferred into nontransgenic naïve recipient mice. These animals were then infected with LCMV and analyzed at various times after infection. At day 8, a subset of P14 transgenic donor cells had induced CD8αα, whereas the bulk of primary P14 responder cells remained CD8αα negative (Fig. 3A, day 8). At day 10 after infection (at the beginning of the contraction phase of the response), the proportion of CD8αα+ P14 effector cells remained unchanged, whereas the CD8αα P14 T cells decreased markedly (Fig. 3A, day 10). Similar to wild-type LCMV-specific responder T cells, CD8αα+ P14 cells also expressed elevated levels of IL-7R (Fig. 3A). CD8αα expression disappeared by day 15, however, whereas IL-7R remained high on the surviving P14 cells (Fig. 3A, day 15). The selective survival of CD8αα-expressing primary effector T cells in these experiments was consistent with their possible role as memory precursors (19).

Fig. 3.

Direct progeny of CD8αα+ effector cells survive and differentiate into memory T cells. (A) Splenocytes from GP33-41–specific TCR transgenic mice (P14; 106 cells per mouse) were transferred into C57BL/6 mice, which were infected with LCMV 1 day later. Splenocytes from these animals were analyzed on days 8, 10, and 15 after infection. After gating on cells stained for Vα2+/Vβ8.1+, T cells were analyzed for CD8αα and IL-7Rα expression. Data represent one of four mice per time point. (B) GP33tet+/CD8αα+DP and GP33tet+SP sorted primary effector P14 cells (Ly5.2+) were transferred a second time into naïve wild-type (Ly5.1) mice (106 cells per mouse). Secondary recipient mice were rested for 40 days and 3 days after rechallenge with LCMV (∼105 plaque-forming units per mouse) and analyzed for the recovery and function of Ly5.2+ splenocytes. Upper panels: recovered Ly5.2+ CD8 cells from one recipient mouse in each case. Middle panels: IFN-γ secretion of these cells stimulated with GP33-41 peptide in vitro. Bar graphs: numbers of recovered P14 cells and IFN-γ–secreting secondary P14 cells from three recipient mice. Data in (B) are from one of two independent experiments.

We next examined the long-term survival of the CD8αα+ CD8αβ+ T cells and their potential to mature toward memory T cells (16). P14 T cells transferred to recipient mice were isolated on day 8 after LCMV infection and sorted into CD8αα+ and CD8αα fractions. The two subsets were then retransferred to naïve recipient animals. To control for potential migration differences between the two cell subsets, we analyzed a set of recipient mice in each case 2 days after cell transfer for the presence of the donor cells in various tissues. Comparable numbers of both cell types were isolated from each tissue analyzed, indicating that CD8αα+ and CD8αα effector cells migrated similarly. Remaining recipients were rested for 40 days and analyzed after rechallenge with the virus. Substantial numbers of P14 T cells were recovered from mice that had received CD8αα+ primary P14 effector cells, but not from those that had received the CD8αα fraction (Fig. 3B). To test for functional differentiation of the surviving donor cells, we measured cytokine responses after in vitro stimulation with GP33-41 peptide. Most progeny of CD8αα+ P14 effector cells displayed a potent secondary response, as measured by intracellular interferon-γ (IFN-γ) staining, whereas CD8αα progeny did not (Fig. 3B).

As a direct test for a role of CD8αα in memory formation, mice carrying an enhancer deletion within the murine cd8 locus (E8I–/–) (20) were examined. This enhancer is required for expression of CD8αα, and although cells from these mice express normal levels of CD8αβ (20), no CD8αα could be detected after stimulation of E8I–/– T cells in vitro (fig. S3). These observations indicated that the induction of CD8αα on activated CD8αβ+ splenocytes is controlled by the E8I enhancer. Consistent with this idea, LCMV-infected E8I–/– mice generated substantial numbers of primary LCMV-specific CD8αβ+ effector cells (Fig. 4, A and B), none of which expressed CD8αα (Fig. 4A). Analysis of E8I–/– and wild-type mice for primary responses revealed comparable amounts of IFN-γ–producing primary T cells from both mice (Fig. 4C). In contrast, the low numbers of E8I–/– LCMV-specific T cells isolated 50 days after the initial infection, and the poor response of these cells upon restimulation in vitro, indicated a severe defect of the E8I–/– lymphocytes in terms of survival and memory cell differentiation (Fig. 4, B and C). Additionally, IL-7Rhigh cells were also absent among primary E8I–/– effector cells, indicating that CD8αα is required for the initial up-regulation of IL-7R on CD8 memory T cell precursors (Fig. 4D).

Fig. 4.

CD8α enhancer–deficient mice (E8I–/–) are compromised in generating memory CD8 T cells. Wild-type (WT) and E8I–/– mice infected with LCMV were analyzed for primary, memory, and secondary response on days 7, 50, and 56 after rechallenge, respectively. (A) Gated NP396 tet+, GP276 tet+, or GP33 tet+ CD8 primary effector T cells analyzed for CD8αα. Data are from one of four mice in each group. (B) Splenocytes stained for GP33/Db tetramer on gated CD8+ cells were assessed for LCMV-specific primary and secondary responses. (C) Dot plots of IFN-γ production from primary, memory, and secondary responder cells measured after in vitro restimulation with GP33-41 peptide. Values indicate percentage of IFN-γ+ cells in total splenocytes. Bar graphs show percentages of IFN-γ + primary, memory, or secondary LCMV effector cells from WT or E8I–/– mice and stimulated in vitro with GP33-41 or NP396-404 peptide. Data in (B) and (C) are from 9 to 12 mice each. (D) LCMV-specific primary effector cells gated on GP33/Db tetramer+ cells were stained for CD8αα and IL-7Rα expression. Data are from one of five mice.

Unique features of CD8αα distinguish this molecule from the conventional TCR coreceptor CD8αβ. CD8β facilitates translocation of the CD8 coreceptor into lipid rafts and localization of the CD8α-associated p56lck in proximity to the TCR activation complex (21). CD8αα is largely excluded from these rafts (21) and could thus sequester p56lck away from the TCR. Dual expression of CD8αα with CD8αβ also appears to mediate raft disruption and an overall decrease of TCR activation signals (22). However, the effect of CD8αα may not simply be inhibitory, as CD8αα might redirect the p56lck toward other immune receptors outside rafts, including IL-7R and IL-15R.

The impaired CD8 T cell memory observed in E8I–/– mice provides direct evidence that CD8αα is a key player in the initial survival of primary effector cells as well as ultimately in memory differentiation itself. The activation-induced expression of TL on APCs at the same time as CD8αα on T cells suggests that CD8αα is likely to mediate its function through interaction with TL. TL is not a typical peptide-binding MHC molecule (23), and the preferential interaction of TL with CD8αα (4) could potentially have the effect of further reinforcing the sequestration of p56lck away from the TCR.

Although at this stage the CD8αα-dependent mechanism for memory formation described here appears to be confined to CD8αβ+ T cells, the previous observations that under certain activation conditions CD4 T cells can induce CD8αα (24, 25) leave open the possibility that a similar process might also operate in CD4 T cell memory. Human CD8αβ+ TCRαβ+ peripheral blood lymphocytes also express CD8αα, and human CD8αα+ T cells with a memory phenotype have recently been identified (13), suggesting that CD8αα-dependent memory formation may be preserved across species.

In the design of vaccines, much attention has been focused on the magnitude of the initial immune effector cell burst in terms of establishing robust immune memory (26). Our results indicate that this is not the sole determining factor and that selective survival and differentiation of CD8αα+ precursors, leading to long-lived CD8 memory T cells, also plays an important role.

Supporting Online Material

Materials and Methods

Figs. S1 to S3


References and Notes

View Abstract

Navigate This Article