Self-reactive memory-phenotype CD8 T cells exhibit both MHC-restricted and non-MHC-restricted cytotoxicity: a role for the T-cell receptor and natural killer cell receptors

In this study, we used mice that express a transgenic T-cell receptor (TCR) that is specific for the male (H-Y) antigen presented by H-2D^b as a model system¹  for studying the developmental requirements and function of a population of self-specific CD8⁺ T cells. H-Y TCR transgenic mice have been widely used as an animal model system for the determination of mechanisms of positive and negative selection of T cells. In these mice, CD8 T cells that express the H-Y TCR are positively selected in the thymus of female H-2^b mice.^1,2  In male H-2^b mice, the H-Y TCR is negatively selected, leading to the massive deletion of double-positive (DP) thymocytes.³  Interestingly, there is a population of CD8⁺ T cells, which express a low level of CD8 (referred to as CD8^lo), that are resistant to deletion in male mice.^4,5  The CD8^lo cells are self-reactive, as they expand after transfer into male but not female H-2^b nude mice.⁵  These cells express high levels of CD44⁵  and interleukin-2 receptor β (IL-2Rβ; CD122)⁶  that are characteristic of memory T cells. However, they differ from conventional memory T cells in that they are more refractory to activation by antigen compared with naive T cells.^7,8  von Boehmer et al⁵  have shown that the development of CD8^lo cells is thymus dependent, since the cells are not found in male H-2^b H-Y nude mice. By contrast, Yamada et al⁶  found that CD8^lo cells developed normally in thymectomized male but not female H-2^b mice. Another unusual feature of these CD8^lo T cells is that they proliferate in response to cytokines such as IL-2 and IL-15 in an antigen-independent manner.⁹  However, the CD8^lo cells are similar to conventional memory T cells with regard to the ability to rapidly produce interferon γ (IFN-γ) upon TCR stimulation¹⁰  and in respect to the killing of susceptible target cells without the need for additional reactivation with antigen.^11,12 

CD8⁺ T cells that develop via an extrathymic pathway and possessing similar functional characteristics are also found in normal (non-TCR transgenic) mice.⁸  More importantly, cells of this phenotype in normal mice have also been shown to be important in responses toward viral and bacterial infections, suggesting that they may have an important role in the immune response.^13,14  We have recently showed that CD8⁺CD44^hi T cells from normal mice possess characteristics of both T cells and NK cells.¹⁵  These cells are activated in response to IL-2 alone and show self-reactivity in that they preferentially kill syngeneic tumor cells. Furthermore, CD8⁺CD44^hi T cells express NK receptors upon activation, and engagement of these NK receptors enhances the ability of the cells to lyse syngeneic tumor cells. In this report we show that CD8⁺ T cells from antigen-expressing H-Y TCR transgenic mice also exhibit characteristics of both T cells and NK cells and are remarkably similar to CD8⁺CD44^hi cells from normal mice with regard to their cell surface and functional phenotype. Thus, the H-Y TCR transgenic mice provide a well-defined system for characterizing the developmental biology and function of this interesting cell type.

Breeders for C57BL/6 (B6), B6-Tap-1^-/-, and DBA/2 were obtained from the Jackson Laboratories (Bar Harbor, ME). BDF₁ mice were F₁ mice from the mating of C57BL/6 mice with DBA/2 mice. The H-Y TCR transgenic (tg) mice were bred to the B6 background. Mice 8 to 12 weeks of age were used for the experiments described.

The following monoclonal antibodies (mAbs) were used: CD4 (GK1.5), CD8α (53-6.7), CD8β (53.58), CD3ϵ (2C11), CD44 (PGP1), H-Y TCRβ (F23.1), H-Y TCRα (T3.70), CD94 (18D3), NK-1.1 (PK136), CD122 (TM-β1), Ly6C (AL-21), CD244.2 (2B4), CD16/32, IL-7Rα (A7R34), NKG2D (A10 and CX5), and anti-NKG2D.¹⁶  Biotinylated mAbs were detected using streptavidin-phycoerythrin (PE). CD8α, H-Y TCRα, NKG2D (A10 and CX5), IL-7Rα, and CD94 were obtained from eBioscience (San Diego, CA). All other antibodies (Abs) were obtained from BD PharMingen (San Diego, CA), except anti-NKG2D,¹⁶  which was a kind gift from Dr Wayne M. Yokoyama (Howard Hughes Medical Institute, Washington University, St Louis). Cell staining and flow cytometry were performed according to standard procedures. The CellQuest software program (Becton Dickinson, Mountain View, CA) was used for data acquisition and analysis.

Cell lines used were the RMA lymphoma (H-2^b+, Rae-1δ^-), RMA-Rae-1δ transfectant (H-2^b+, Rae-1δ⁺), and P815 mastocytoma. The cell lines were cultured in Iscove modified Eagle medium (Life Technologies, Burlington, ON, Canada), supplemented with 10% (vol/vol) fetal bovine serum (FBS; Life Technologies), 5 × 10⁵ μM 2-mercaptoethanol (2-ME), and antibiotics (I-medium). The RMA-Rae-1δ transfectant¹⁷  was a kind gift from Dr Lewis L. Lanier (University of California, San Francisco).

Single-cell suspensions from the lymph nodes (LNs) of mice were treated with biotinylated anti-CD8β mAb followed by positive selection using the MiniMACS system (Miltenyi Biotech, Auburn, CA), according to the manufacturer's specifications. The resulting cells were more than 95% pure CD8αβ⁺TCRβ⁺ T cells. For purification of H-Y TCRα (T3.70)⁺ CD8⁺ T cells, single-cell suspensions from the lymph nodes (LNs) of H-Y male and female mice were treated with anti-CD4 mAb and then depleted of CD4⁺ immunoglobulin-positive (Ig⁺) cells using Dynabeads M-450 Sheep anti-mouse IgG (Dynal Biotech, Lake Success, NY), according to the manufacturer's instructions. After depletion, the cells were stained with anti-CD8β-fluorescein isothiocyanate and T3.70-PE and then sorted on a Becton Dickinson fluorescence-activated cell-sorter (FACS) Vantage SE Turbo sort cell sorter. Cell sorting was performed by Andrew Johnson (University of British Columbia, Vancouver, BC, Canada), and the resulting cells were more than 95% pure.

Natural killer (NK) cells were purified and activated as previously described.¹⁵ 

Purified CD8⁺ T cells (1 × 10⁷/mL) were labeled with 1 μM carboxyfluorescein succinimidyl ester (CFSE; Molecular Probes, Eugene, OR) in phosphate-buffered saline for 8 minutes at room temperature. After stopping the reaction with the addition of an equal volume of FBS, cells were washed 4 times with complete media prior to use.

Purified CFSE-labeled CD8 T cells (1-2 × 10⁶) from H-2^b/b male and female and H-2^b/d male H-Y mice were injected into male or female B6 mice that had received a sublethal dose of irradiation (650 cGy) 24 hours earlier. On day 7, spleens of the injected mice were depleted of CD4⁺Ig⁺ cells and stained with the indicated mAbs and analyzed by FACS. For Listeria monocytogenes (LM) infections, purified H-Y CD8 T cells (3 × 10⁶) were labeled with CFSE and transferred into B6 male or female mice. Then, 24 hours later the mice were challenged with LM (10 000 colony-forming units). At 5 days after infection, the mice were killed and the spleens of the animals were analyzed. For the analysis of IFNγ production, 5 × 10⁶ splenocytes from infected and uninfected mice (day 5) were cultured in 1 mL I-medium containing GolgiStop (BD PharMingen) in the presence or absence of H-Y peptide (1 μM). After 6 hours of incubation, the cells were fixed, permeabilized, and stained with anti-IFNγ, anti-CD8, and anti-H-Y TCR.

Purified H-Y CD8 T cells (10⁴) were cultured with irradiated B6-Tap-1^-/- splenocytes (5 × 10⁵) and the indicated concentration of H-Y peptide (KCSRNRQYL)¹⁸  in the presence or absence of IL-2 (20 U/mL). The cells were pulsed with 1 μCi (0.037 MBq) ³H-thymidine for the final 6 hours of a 72-hour culture. For CFSE-based proliferation assays, 1 × 10⁵ CFSE-labeled CD8 T cells were cultured in 96-well round-bottom plates in the presence of IL-2 (200 U/mL), IL-15 (100 ng/mL) or Tap-1^-/- splenocytes (10⁶), H-Y peptide (1 μM), and IL-2 (20 U/mL). CFSE dilution (cell division) was assessed by FACS at the indicated times.

Cytotoxic T lymphocyte (CTL) assays were performed as previously described.¹⁵  The CTL activity of activated CD8 T cells against RMA, RMA-Rae-1δ target cells was assessed at a ratio of 10 effector T cells to 1 target cell in a 4-hour ⁵¹Cr-release assay. For redirected lysis experiments, day-4 activated CD8 T cells were preincubated with the indicated mAb (10 μg/mL) for 15 minutes. The CTL activity of Ab-coated CD8⁺ T cells against Fc receptor (FcR⁺) P815 target cells was then determined in a 4- to 5-hour ⁵¹Cr-release assay. Spontaneous release varied from 8% to 15% of the maximum. All assays were performed in triplicate. Percent specific lysis was calculated as 100% × (cpm [experimental] - cpm [spontaneous release])/(cpm [maximum release] - cpm [spontaneous release]).

Sorted H-Y female T3.70⁺CD8⁺ cells and male T3.70⁺CD8⁺ cells (10⁶) were activated with Tap-1^-/- stimulators (10⁷), H-Y peptide (1 μM), and IL-2 (20 U/mL) for 4 days. NK cells were activated with IL-2 (200 U/mL) for 5 days. RNA isolation and PCR primers were described previously.¹⁵ 

Purified CD8 T cells were stimulated for 10 minutes at 37°C with anti-CD3ϵ (2C11) or with phorbol myristate acetate (PMA, 25 ng/mL) plus ionomycin (500 ng/mL) and then pelleted and lysed in 10 mM Tris (tris(hydroxymethyl)aminomethane, pH 7.5), 150 mM NaCl, 1% Triton X-100, 0.1% sodium dodecyl sulfate, and protease and phosphatase inhibitors. The lysates were separated on a 4% to 15% Tris-HCl polyacryl-amide gel and transferred to polyvinylidene fluoride membrane. Blots were developed using enhanced chemiluminescence system (Amersham, Baie d'Urfé, Québec, Canada). Phospho-zeta-associated protein 70 (ZAP-70), ZAP-70, and phospho-extracellular signal-related kinase 1/2 (ERK1/2) antibodies used for detection were from Cell Signaling Technologies (Beverly, MA). Anti-ERK mAb was from Santa Cruz Biotechnology (Santa Cruz, CA).

Student t test was used to determine P values in CTL assays.

The H-Y TCR is specific for a male peptide presented by H-2D^b.¹ Figure 1A shows the CD4/CD8 thymocyte profiles of H-2^b/b H-Y male and female mice as well as H-2^b/d H-Y male mice. It is clear that the presence of cognate antigen in H-2^b/b male mice results in the deletion of the vast majority of DP thymocytes. In H-2^b/d H-Y male mice, where there is half the number of the deleting H-Y/D^b complexes, there is incomplete deletion of DP thymocytes. Deletion of DP thymocytes in H-2^b/b and H-2^b/d male mice results in drastically reduced numbers of CD8 single-positive (SP) thymocytes relative to H-2^b H-Y female mice (Figure 1A).

Although the majority of H-Y TCR⁺ CD8 T cells are deleted in the male H-2^b/b and H-2^b/d mice, there is a large population of CD8⁺ H-Y TCR⁺ T cells in the spleen and lymph nodes of these mice. However, CD8⁺ T cells from male H-Y mice differ from CD8⁺ T cells from female H-Y mice in many aspects (Figure 1B). Female H-2^b/b CD8⁺ T cells express the highest level of the CD8 coreceptor (referred to as CD8^hi); those from H-2^b/d male mice express an intermediate level (referred to as CD8^int); and those from H-2^b/b male cells express the lowest level (referred to as CD8^lo). There is also a correlation between the degree of cognate self-antigen exposure and the level of expression of memory markers in male H-Y mice. In female mice, the lack of cognate self-antigen exposure results in CD8⁺ T cells with a naive phenotype (CD44^lo/int, IL-2Rβ^-, IL-7Rα^lo, Ly6C^-). In contrast, the CD8⁺ T cells from H-Y male mice exhibit an activated/memory phenotype (CD44^int/hi, IL-2Rβ⁺, IL-7Rα^hi, Ly6C⁺). Notably, the CD8^lo cells expressed higher levels of memory markers compared with CD8^int cells.

Previous studies show that the CD8^hi cells from female H-Y TCR transgenic mice can express both the transgenic as well as an endogenous α chain.¹⁹  In stark contrast, the CD8 T cells from H-Y male mice express exclusively the transgenic TCRα chain (Figure 1B). Using the TCRα expression as a measure of TCR expression, it is clear that the CD8^lo and CD8^int cells expressed the same intermediate level of the αβ TCR compared with CD8^hi T cells, which express a high level (Figure 1B). Corroborating results were obtained by analyzing expression of TCRβ and CD3ϵ on these cells (data not shown).

In agreement with a previous study,⁸  we found that the CD8^lo cells proliferated poorly to antigen stimulation in the absence of IL-2 (Figure 2A, upper panel). The CD8^int cells also showed a similar proliferative defect. This proliferative defect was partially restored by exogenous IL-2 (Figure 2A, lower panel).

The increased activation threshold of the CD8^int and CD8^lo cells may be due to a defect in TCR signal transduction. Signaling through the TCR leads to phosphorylation of CD3ζ, leading to the recruitment and subsequent phosphorylation of ZAP-70.²⁰  ZAP-70 activation eventually leads to the activation of several important signaling pathways including the Ras/mitogen-activated protein kinase (MAPK) pathway. Figure 2B clearly shows that CD8^int and CD8^lo cells have a defect in early TCR signaling, as they do not efficiently phosphorylate ZAP-70 in response to TCR stimulation compared with conventional, naive CD8^hi cells. Correlated with diminished early TCR signaling, the cells also have a defect in downstream signaling events, leading to reduced p42/p44 (ERK1/2) MAPK phosphorylation. The inefficient activation of ZAP-70 correlates with the amount of cognate self-antigen the cells have encountered in vivo; the naive phenotype CD8^hi cells exhibit the strongest signaling, followed by the CD8^int and then the CD8^lo cells. This is in contrast to true memory CD8 T cells, which have a lower activation threshold and more efficiently induce the phosphorylation of ZAP-70 and ERK1/2 relative to naive cells.²¹  Interestingly, ERK1/2 cells were efficiently phosphorylated in response to PMA and ionomycin (Figure 2B), suggesting that there is no intrinsic defect in the ability of these cells to activate the ERK-MAPK pathway.

Since the CD8^lo and CD8^int cells are recovered from mice that express different levels of cognate self-antigen, they offer an opportunity for determining whether prior antigen history affects their responsiveness to inflammatory cytokines. This was determined by culturing CFSE-labeled CD8^lo and CD8^int cells with an exogenous source of IL-2 or IL-15. Proliferation at 48, 72, and 96 hours was determined by measuring the CFSE fluorescence level of the cultured CD8⁺ cells. CD8^hi cells from female H-Y mice do not proliferate in response to IL-2 or IL-15,⁹  so we stimulated these cells with antigen + IL-2 to determine their proliferative potential. The results in Figure 3A indicate that stimulation of CD8^hi, CD8^lo, or CD8^int cells with a high antigen dose + IL-2 led to similar extent of cell division at 48, 72, or 96 hours. In contrast, only the CD8^lo and CD8^int cells could proliferate in response to cytokine alone (Figure 3A). Interestingly, CD8^lo cells proliferated much more rapidly than the CD8^int cells in response to IL-2 at the 72- and 96-hour time points (Figure 3A). IL-15 was more efficient than IL-2 in inducing the growth of CD8^lo and CD8^int cells, particularly at the 48- and 72-hour time points. These studies indicate that the prior antigen history of CD8⁺ cells from male H-Y mice determine the qualitative and quantitative aspects of their responsiveness to inflammatory cytokines.

The antigenic requirements for the expansion of CD8⁺ T cells from male H-Y mice in vivo was determined in adoptive transfer experiments where CFSE-labeled CD8^hi, CD8^int, and CD8^lo cells were injected into irradiated female or male B6 mice. Figure 3B indicates that the CD8^int and CD8^lo cells homed to and expanded efficiently in the spleen of either cognate-antigen-expressing male or non-antigen-expressing female mice, whereas the CD8^hi cells could expand only in the presence of antigen. A recent report has also shown that CD8^lo cells can in fact expand in a major histocompatibility complex (MHC) class I-deficient host.²²  In our experiments, we showed that CD8^lo cells could expand in female recipients, and they underwent one extra division in male recipients. CD8^int cells, on the other hand, were much more dependent on the male antigen for proliferation since they underwent about 3 fewer divisions in female recipients. In all cases, the presence of cognate antigen affected the clone sizes, suggesting that antigen does in fact play a role in the lymphopenia-induced expansion of these memory-phenotype CD8 T cells. This expansion in the absence of cognate self-antigen is likely mediated through interaction of IL-2Rβ and or IL-7Rα with IL-15 and IL-7.²³  By contrast, CD8^hi cells could not expand in irradiated female mice as interactions between the H-Y TCR and female self-peptides/H-2D^b alone are insufficient for the expansion of CD8^hi cells in vivo.^22,24  CD8^hi cells proliferated most efficiently in irradiated male mice.

The data in Figure 3 suggest that CD8^int and CD8^lo cells are dependent on interaction with the self (male)-antigen for their development and are likely to be of the same lineage. Since CD8^int cells consistently showed an intermediate phenotype relative to CD8^lo cells, we decided to focus the remainder of our experiments on CD8^lo and CD8^hi cells to further characterize the difference between these 2 cell types. Since CD8^lo cells can expand in an antigen-independent manner in irradiated female mice, we determined if these cells could undergo bystander proliferation in response to bacterial infection. To this end, we transferred CD8^lo cells into nonirradiated B6 female and male mice and then infected the mice with a sublethal dose of LM. As a control, we also transferred CD8^hi cells into nonirradiated B6 female mice and infected them with LM. Figure 4 shows that CD8^lo cells expand in both male and female recipients during infection, although this expansion was greater in male mice. The CD8^hi cells did not expand in either infected or uninfected recipients, and in fact their numbers were depleted in the infected recipients. Consistent with this, a previous study has shown that nonspecific T cells are depleted early during infection with Listeria.²⁵  Our results show that rather than being depleted, CD8^lo cells expand in infected mice, suggesting that inflammatory cytokines produced in response to infection could promote the growth of these cells in vivo.

We recently showed that memory phenotype CD8⁺CD44^hi cells from normal mice express NK receptors after activation.¹⁵  Since CD8^lo cells are similar to CD8⁺CD44^hi cells in terms of cell surface phenotype and cytokine responsiveness, we determined whether they also shared the ability to express NK receptors after activation. Immediately ex vivo, CD8^hi and CD8^lo cells are negative for all of the NK receptors tested, with the exception of a very low level of NKG2D and CD94 expression by the CD8^lo cells (Figure 5A). Since both CD8^hi and CD8^lo cells can be activated with antigen + IL-2, this protocol was used to activate both populations. Figure 5B shows that activated CD8^lo T cells expressed several NK receptors, especially 2B4, DX5, and CD94. By contrast, activated female CD8^hi cells do not express DX5 or CD94. A high level of 2B4 was expressed by some of the activated female CD8^hi cells. NKG2D is known to be expressed by all activated CD8 T cells²⁶  and was expressed equivalently by both activated CD8^hi and CD8^lo cells. Neither cell type expressed the B6 NK-T-cell marker NK1.1.

Conventional antigen-activated CD8 T cells express NKG2D in association with the DNAX activation protein 10 (DAP10) adaptor molecule.^27,28  By contrast, NKG2D in activated murine NK cells is associated with both the DAP10 and DAP12 adaptor molecules.^27,28  Association of NKG2D with DAP10 provides a costimulatory signal, whereas association with DAP12 confers a directly stimulatory signal.²⁷  We recently showed that CD8⁺CD44^hi cells from normal mice expressed both DAP10 and DAP12 after IL-2 activation.¹⁵ Figure 5C showed that activated CD8^lo cells from male mice also expressed both DAP10 and DAP12, whereas antigen-activated CD8^hi cells from female mice expressed only DAP10. Similar results were obtained following activation of CD8^lo cells with IL-2 (data not shown).

We next determined if NKG2D could participate in the lysis of target cells expressing self-antigen. Activated CD8^hi and CD8^lo cells were used as effector cells in a CTL assay against peptide-pulsed RMA (H-2^b) cells or RMA cells transfected with Rae-1δ (RMA-Rae-1),¹⁷  a ligand for NKG2D.²⁶  It is clear that the NKG2D receptor participated in the lysis of RMA target cells by activated CD8^lo T cells since RMA-Rae-1 target cells were lysed significantly better than RMA target cells (Figure 6A). This is particularly evident at low H-Y peptide concentrations, where additive effects between the TCR and NKG2D in the lysis of RMA-Rae-1 target cells were observed. However, at high concentrations of antigenic peptide, RMA and RMA-Rae-1 target cells were killed to the same extent. This result indicates that NKG2D contributes to killing when there is suboptimal stimulation of the TCR by the antigenic ligand. By contrast, there was no difference in the lysis of RMA and RMA-Rae-1 target cells by activated female CD8^hi cells at all concentrations of added H-Y peptide, consistent with the finding that NKG2D is not an activating receptor in conventional CD8 T cells.

We used a redirected killing assay to provide further evidence that NKG2D is an activating receptor in activated CD8^lo cells. In this assay, killing of an Fc receptor-expressing target (P815; H-2^d) ± an anti-NKG2D mAb was determined. We found that only anti-CD3ϵ, but not anti-NKG2D, was able to induce lysis of P815 target cells by activated CD8^hi cells (Figure 6B). This is consistent with the conclusion that NKG2D does not function as an activating receptor in conventional CD8 T cells.^27,28  By contrast, anti-NKG2D could enhance the lysis of P815 target cells by CD8^lo cells in the redirected killing assay (Figure 6B). It is noted that activated CD8^lo cells exhibited significant killing of P815 targets even in the absence of any added antibody. This background killing of P815 targets was not due to NKG2D since we could not eliminate this background killing by using a blocking anti-NKG2D mAb (data not shown). It is conceivable that the non-antibody-dependent killing of P815 targets by activated CD8^lo cells may be due to P815 ligands, which are recognized by other activating NK receptors.

Since CD8^lo cells could expand in response to LM infection (Figure 4), we determined if LM infection induces expression of NKG2D on these cells. In addition, we wanted to determine the effect of bacterial infection on IFNγ production by CD8^lo cells. We transferred CFSE-labeled CD8^lo cells into nonirradiated B6 male mice and then infected them with LM. It is clear that NKG2D was expressed by the majority of cells that have divided in response to infection (Figure 7A). Furthermore, cells that have undergone more cell divisions expressed the highest level of NKG2D. To test the production of IFNγ by the cells, we cultured the splenocytes from infected or uninfected mice for 6 hours in the presence or absence of H-Y peptide. The CD8^lo cells from both infected and uninfected mice did not produce any IFNγ without any TCR stimulation (Figure 7B). However, upon a 6-hour stimulation with the antigenic peptide, CD8^lo cells from uninfected and LM-infected mice produced high levels of IFNγ. By contrast, CD8^hi cells from uninfected or infected recipients did not produce any IFNγ in response to similar stimulation (data not shown). CD8^lo cells from LM-infected mice produced higher levels of IFNγ than those from uninfected mice. Furthermore, CD8^lo cells that have undergone the most number of cell divisions in response to LM infection produced the highest level of IFNγ (Figure 7B). Our observation that CD8^lo cells from uninfected mice can produce IFNγ in response to antigen stimulation is consistent with the observation that CD8^lo cells can produce IFNγ in response to TCR stimulation without the need for prior activation.⁸ 

In this report we have characterized an unusual population of self-antigen-specific CD8⁺ T cells in H-Y TCR transgenic mice. The development of these cells is dependent on interaction with cognate self-antigen, and the intensity of interaction with cognate-antigen in vivo determines the expression level of CD8 and memory markers. However, in contrast to conventional memory cells, these cells have an increased threshold for activation relative to naive CD8⁺ T cells of the same antigen specificity. They proliferate in response to cytokines such as IL-2 and IL-15. When introduced into lymphopenic mice, these cells can expand in the absence of self-antigen. More interestingly, they proliferate in response to a bystander bacterial infection in normal mice. Immediately ex vivo, these cells are negative or express low levels of NK receptors, but upon activation several NK receptors are induced, including NKG2D. Interestingly, activated male H-Y CD8^lo cells express DAP12, an adaptor molecule normally expressed by NK cells.^27,28  Furthermore, the NKG2D receptor acts additively and also functions independently of the TCR in the killing of NKG2D ligand-positive cells. We have recently reported the existence in normal mice of CD8⁺CD44^hi T cells that possess cell surface phenotypes and functional properties similar to those in the male H-Y CD8^lo T cells.¹⁵  These results support the hypothesis that the H-Y CD8^lo cells are not a transgenic oddity but are a normal component of the murine immune system.

The H-Y TCR transgenic mouse provides a well-defined system for studying the developmental biology of this novel cell type. In this study we showed that they could be selected by a conventional antigen, that is, the male peptide presented by MHC class Ia molecules. We also showed that the expression level of CD8 on this cell type is variable and reflects their prior antigen history. The degree of TCR/cognate self-antigen interaction in vivo also affects the response of these cells to IL-2 or IL-15 in vitro (Figure 3). Proliferation in response to IL-2 occurred much more efficiently in CD8^lo cells than CD8^int cells, presumably as a result of higher IL-2Rβ expression (Figure 1).

CD4⁺ or CD4^-CD8^- NKT cells bearing an invariant TCRα chain (Vα14 in mice and Vα24 in humans) that are reactive to CD1d also share many similarities with CD8^lo cells. NKT cells are an important component of the innate immune response to infection and can produce cytokines to activate other cells or can directly lyse some targets (reviewed in Kronenberg and Gapin²⁹ ). Although these NKT cells react strongly to α-galactosyl-ceramide (a sphingolipid from a marine sponge) presented by CD1d, it was recently shown that their activation in response to infection is dependent on CD1d/self-antigen and IL-12.³⁰  Both CD8^lo cells and NKT cells exhibit an activated/memory phenotype and are thought to be self-reactive.³¹  Another striking similarity is that NKT cells are selected by high-affinity interactions with self-antigen on cells other than thymic cortical epithelial cells,³²  and this selection can occur in the absence of a thymus.³³  MHC class Ib-restricted CD8 T cells are also thought to be selected by relatively high-affinity interactions with self-antigens³⁴  and are also functionally similar to CD8^lo and NKT cells in that they mount an early response against bacterial infection.³⁵  It is tempting to speculate that cells such as NKT, and MHC class Ib-restricted CD8⁺ and CD8^lo cells, which are selected by high-affinity interactions with self-antigens, may be members of a family of T cells that evolved to provide an early defense mechanism against bacterial infection.

Our data showed that NK receptors such as NKG2D can function as an activating receptor in male CD8^lo cells but not in female CD8^hi cells. NKG2D has recently been shown to be important for the recognition and destruction of tumor cells by NK cells.^17,26,36  In vivo, the expression of the ligands for NKG2D, Rae-1, and H60 results in the rapid clearance of ligand-positive tumors with the generation of protective immunity against subsequent challenge with parental, ligand-negative tumors.³⁶  NKG2D has been demonstrated to associate with 2 different adaptor molecules, DAP10 and DAP12.^27,28  In CD8⁺ T cells, which express only DAP10, NKG2D engagement results in a costimulatory signal, whereas in activated NK cells, the expression of DAP12 allows NKG2D to provide an activating signal.^27,28  In addition, it has been shown that the ectopic expression of DAP12 in CD8 T cells results in the ability of NKG2D to transduce a directly stimulatory signal.²⁸  Interestingly, we have found that activated CD8^lo cells express both DAP10 and DAP12, unlike CD8^hi cells, making CD8^lo cells similar to NK cells in this regard.

Despite the fact that H-Y CD8^lo cells are self-reactive, the presence of a large number of these self-reactive cells in male H-2^b H-Y mice did not lead to autoimmune disease. It is conceivable that the increased activation threshold of these cells is sufficient to prevent autoimmunity. We have shown that CD8^lo and CD8^int cells have a defect in TCR signaling, which can be partially compensated for by the addition of exogenous IL-2. This defect stems from the inability of CD8^int and CD8^lo cells to efficiently phosphorylate signaling proteins such as ZAP-70 and ERK, in response to TCR stimulation (Figure 2B). A recent study supports the hypothesis that the lowering of the T-cell activation threshold can lead to autoimmune diseases in male H-Y TCR transgenic mice. Murga et al found that T cells with a null mutation in the E2F2 transcription factor have a lowered activation threshold.³⁷  Interestingly, male H-2^b H-Y TCR transgenic mice with this null mutation develop an accelerated and much more severe lupuslike autoimmune syndrome than normal mice with this mutation. The unique combination of high activation threshold, the ability to respond to inflammatory cytokines such as IL-15 and undergo bystander proliferation, and the expression of activating NKG2D receptor in response to activation would render these self-specific CD8⁺ T cells particularly adept at sensing infected and transformed cells. The observed cooperation between the TCR and NKG2D in the destruction of Rae-1⁺ target cells would also allow these cells to focus on normal cells that express stress ligands in response to infection or transformation. These cells would also differ from NK cells in their function, since the expression of MHC class Ia antigens on target cells does not inhibit their function. For the CD8^lo cells, MHC class Ia could in fact serve as an activating antigen. Thus, infected or transformed cells that are not susceptible to killing by NK cells, by virtue of high MHC class Ia expression, would be susceptible to lysis by activated self-specific CD8⁺ T cells. The fact that a bystander infection results in the proliferation of CD8^lo as well as the up-regulation of NKG2D suggests that these cells could use this receptor to eliminate infected cells early during an infection. This non-MHC-dependent function of CD8^lo cells broadens the range of target cells and increases their versatility in surveillance against infected and transformed cells.

We thank Dr Lewis Lanier (University of California, San Francisco) for providing us with the RMA-Rae-1δ cell line. We also thank Dr Wayne Yokoyama for providing the anti-NKG2D mAb and Dr Hao Shen (University of Pennsylvania) for the gift of wild-type LM strain 10403s.