CD1d-Restricted T Cells License B Cells to Generate Long-Lasting Cytotoxic Antitumor Immunity In vivo

Of the available vaccine approaches, cellular vaccines using antigen-presenting cells (APC), such as dendritic cells (DC), are reliable at generating effective T-cell immunity (1). DCs are ideal APCs for immunotherapy because they can capture antigen and then migrate into lymphoid organs where they present the antigen to the relevant T cell. More importantly, they provide strong costimulation to the T cells (2, 3). The DC-based vaccine approach is well established for both experimental and clinical studies. However, DCs are relatively scarce in blood and lymphoid tissues, and it is difficult to increase their numbers ex vivo from blood monocytes, both of which present major drawbacks to their widespread use in vaccines (4).

B cells offer an attractive alternate source for cellular vaccines in that they are abundant in lymphoid tissues and blood (4), easily expanded ex vivo (5, 6), and home to lymphoid organs after parenteral administration. Despite these advantages, B cells have been ignored as a vaccinating APC because they are poorly immunogenic. In fact, accumulating evidence shows that they induce T-cell tolerance in both CD4 and CD8 T cells directly, probably due to the lack of costimulation (7–9). However, ‘activated’ B cells can prime both CD4 and CD8 T cells (5, 6, 10–13), suggesting that, when activated by the appropriate stimuli, B cells can act as immunogenic APCs capable of inducing antigen-specific T-cell immunity.

It is well established that iNKT cells play a crucial role in a variety of immune responses and in immunopathology as a whole (14, 15). They act as regulators of immunity in tumor (16), diabetes (17, 18), and at immune-privileged site (19). In sharp contrast, ligand-activated iNKT cells lead to the activation of T, B, and natural killer (NK) cells as well as DCs. Injection of α-galactosylceramide (αGalCer), an iNKT ligand, generates antitumor immunity via the mediation of NK and T cells (20). Mice, to which protein antigen and αGalCer have been cogiven, develop humoral and cell-mediated immunity, including CTL responses (21, 22). Interestingly, Crowe et al. (23) has recently reported that CD4⁻ iNKT cells in liver could induce better antitumor immunity than CD4⁺ iNKT in liver or iNKT from other organs. Furthermore, Terabe et al. (24) showed that type II NKT population negatively regulates antitumor immunity. Thus, it is likely that respective NK T-cell (NKT) subsets in different lymphoid organs possess their own novel function in vivo. Moreover, a recent study has shown that αGalCer-loaded DCs generate longer lasting iNKT cell responses than does free form of αGalCer (25). Based on these findings, we hypothesized that presentation of the iNKT ligand on B cells could convert them from tolerogenic to immunogenic, thereby generating strong immunity against antigen displayed on MHC molecules of the B cells. To test this hypothesis, we determined the efficiency of αGalCer-loaded, peptide-pulsed B cells in generating cytotoxic immunity and antitumor activity.

Mice and antibodies for depletion. Female C57BL/6 and BALB/c mice (Charles River, Seoul, Korea) were used at 6 to 10 weeks of age. The OT-I and C57BL/6^bm1 (bm-1) mice were purchased from The Jackson Laboratory (Bar Harbor, ME), whereas Jα281^−/− and MHC II^−/− mice were kindly provided by Dr. Doo-Hyun Chung (Seoul National University, Seoul, Korea) and Dr. Se-Ho Park (Korea University, Seoul, Korea), respectively. All mice were kept under specific pathogen-free conditions in the Animal Center for Pharmaceutical Research (Seoul National University). Antibodies from hybridomas [i.e., GK1.5 (anti-CD4), 2.43 (anti-CD8), and PK136 (anti-NK1.1)] were obtained and injected i.p. (150 μg/mouse) to deplete the respective lymphocyte subsets in vivo. All animal studies were approved by the Institutional Animal Care and Use Committee of Seoul National University.

Loading αGalCer and peptide on B cell or DC. B220⁺ cells were purified from splenocytes using microbeads once CD11c⁺ cells had been depleted by anti-CD11c microbeads. These cells were >99% CD19 positive. DCs were purified from spleen by treatment with collagenase IV and DNase I and by density gradation followed by sorting with CD11c microbeads as described previously (26).

Purified cells (B cells or DCs) were cocultured with αGalCer (1 μg/mL), which had been prepared as described previously (27), or the vehicle (0.5% polysorbate) for 14 hours in a CO₂ incubator. In some experiments, these cells were also additionally cocultured with ovalbumin_257-264 peptide or HER-2/neu_63-71 for 1 hour. After extensive washing, these cells were i.v. injected into their syngenic mice or cocultured with a NKT hybridoma (DN32.D3).

Carboxylfluorescein diacetate succinimidyl ester–labeled OT-I adoptive transfer study. Ovalbumin-specific CD8 T cells (>90% of which were Vα2 positive) were isolated from OT-I mice using magnetic beads. These cells were labeled with 10 μmol/L carboxylfluorescein diacetate succinimidyl ester (CFSE) and i.v. transferred into their syngenic mice (26). On the following day, mice were i.v. injected with B cells manipulated in vitro with indicated conditions. Forty-eight hours later, lymphoid cells from the lymph nodes or spleen of the recipient mice were stained with phycoerythrin (PE)-conjugated anti-Vα2 antibody and then analyzed by flow cytometry.

Intracellular cytokines staining. For detection of intracellular cytokines in OT-I T cells, lymphoid cells were stimulated for 6 hours in RPMI 1640 supplemented with 10% FCS and Golgistop in the presence or absence of 1 μmol/L ovalbumin_257-264. Cells were permeabilized with Cytofix/Cytoperm reagents in accordance with the manufacturer's recommendations before being stained with PE-conjugated anti-IFN-γ monoclonal antibody (mAb) or PE-conjugated anti-CD25 antibody (BD PharMingen, San Diego, CA). For intracellular staining of IFN-γ in iNKT cells, splenocytes from mice injected with vehicle-pulsed B cells (B/veh) or αGalCer-loaded B cells (B/αGalCer) were stained with αGalCer/CD1d-multimer together with PE-Cy5-conjugated anti-T-cell receptor β (TCRβ) antibody. These cells were permeabilized and stained with APC-conjugated anti-IFN-γ antibody.

In vivo and in vitro cytotoxicity assay. The in vivo ovalbumin-specific cytolytic activity of CD8 T-cell responses was measured using flow cytometry. Syngenic lymphocytes were either loaded with 1 μmol/L peptides or left untouched before being labeled with CFSE at different concentrations (20 and 2.5 μmol/L, respectively). Equal numbers of the two populations were mixed and injected i.v. into mice. Eighteen to 24 hours later, lymphoid cells from spleen and lymph nodes were analyzed to assess peptide-specific killing. The ovalbumin-specific lysis was calculated as follows: r = % CFSE^low / % CFSE^high and % lysis = [1 − (r_unprimed / r_primed)] × 100, where r is the ratio. In vitro cytolytic activity against EG-7 target cells was measured using a standard ⁵¹Cr release assay as described previously (28).

Detection of a peptide-specific CTL population. The population of peptide-specific CTL was calculated based on IFN-γ-producing CD8 T cells induced in response to ovalbumin_257-264 (26). Briefly, cells from spleen were stimulated with 1 μmol/L ovalbumin_257-264 for 4 hours in the presence of 1 μg/mL GolgiPlug. Cells were fixed, permeabilized, stained with antibodies to mouse IFN-γ-PE and CD8-FITC, and then analyzed by flow cytometer.

Preventive and therapeutic tumor model. Ovalbumin-transfected B16 melanoma (MO-5, kindly provided by Dr. Kenneth Rock, University of Massachusetts, Worcester, MA) and HER-2/neu-transfected CT26 colon carcinoma (CT26-HER-2/neu; ref. 29) were used as model tumors. To test preventive effects, C57BL/6 mice were given various cellular vaccines at day 0 and then 2 × 10⁵ MO-5 were s.c. injected into their left flank on day 7. To test therapeutic effects, mice received 2 × 10⁵ MO-5 on day 0 and were then given cellular vaccines on days 1 or 9. In the HER-2/neu tumor model, BALB/c mice were challenged i.v. or s.c. with 2 × 10⁵ CT26-HER-2/neu, causing them to develop tumors. The tumor-bearing mice were then vaccinated with B-cell-based or DC-based cellular vaccines, and rates of survival or tumor growth were measured. In some experiments, tumor-free mice were further inoculated s.c. with the same tumor and tumor growth was monitored.

Bidirectional activation of B/αGalCer and iNKT cells in vivo. Because it was already well established that αGalCer-loaded DCs activate iNKT cells (25), we first examined whether B/αGalCer would do likewise. After depleting CD11c⁺ cells from the splenocytes to remove DC contamination, we isolated pure B cells using anti-B220 microbeads. The sorted cells were CD19⁺, CD1d⁺, but CD11c⁻ (>99%; Fig. 1A). Primary DCs were isolated from spleen (26). These purified B cells or DCs were pulsed with various concentrations of αGalCer and then cocultured with a NKT hybridoma, DN32.D3 (30). B/αGalCer efficiently stimulated DN32.D3 cells to produce interleukin (IL)-2 (Supplementary Fig. S1A), equaling the rate of the DC group of IL-2 production when at higher ratios to the hybridoma but falling short when at lower ratios (Supplementary Fig. S1B).

To examine if B/αGalCer could directly induce iNKT activation in vivo, we injected B/αGalCer or B/veh i.v. into syngenic mice and measured the level of intracellular IFN-γ in iNKT cells. Indeed, injection of B/αGalCer stimulated iNKT to produce IFN-γ whereas B/veh did not (Fig. 1B). Consistent with this finding, we observed the induction of IL-4 and IFN-γ-producing cells on B/αGalCer injection in wild-type (WT) mice but not in Jα281^−/− mice, showing the activation of iNKT cells by B/αGalCer in vivo (Supplementary Fig. S1C).

To determine what, if any, changes were induced in B cells after injection, we next i.v. injected CFSE-labeled B/αGalCer into syngenic mice and then analyzed the costimulatory molecules on the CFSE⁺ cells. Flow cytometric analysis of B cells revealed that αGalCer minimally affected the level of CD40, CD80, and CD86 on B cells during the in vitro pulsing period (Supplementary Fig. S1D). Interestingly, high levels of CD86 but not CD80 expression were induced within 24 hours (Fig. 1C). CD40 and MHC II were also slightly up-regulated. Even 48 hours after injection, no up-regulation of CD80 was observed, whereas all other results remained largely consistent with the 24-hour level (data not shown). Therefore, both DC/αGalCer and B/αGalCer are capable of activating iNKT cells both in vitro and in vivo.

Peptide-pulsed B/αGalCer promotes the activation of peptide-specific CD8 T cells. Next, we addressed whether copulsing of αGalCer and MHC I–restricted peptide on B cells could prime peptide-specific CD8 T cells. To this end, we first adoptively transferred CFSE-labeled ovalbumin-specific CD8 T cells (OT-I) into C57BL/6 mice and then i.v. injected B/veh (B alone), B/αGalCer, vehicle plus peptide-pulsed B cells (B/pep), or αGalCer plus peptide-pulsed B cells (B/αGalCer/pep), respectively.

As expected, little division of OT-I cells was induced in mice receiving B alone or B/αGalCer (Fig. 2). Injection of B/pep induced a substantial division of OT-I, but very few of these cells produced IL-2 (<4%) and some population (38%) produced IFN-γ after restimulation ex vivo. By contrast, mice given B/αGalCer/pep showed an enhanced division of OT-I, with >40% of the resultant cells producing IL-2 and, most surprisingly, >90% producing IFN-γ at much higher levels than the B/pep group. These results suggest that a far higher rate of CD8 T-cell activation could be achieved by the loading of αGalCer onto B/pep.

B/αGalCer/pep induces long-lasting cytotoxic T-cell responses. In the next study, we asked if our B-cell-based vaccine approach could induce cytotoxic immunity. To answer this question, we injected i.v. groups of C57BL/6 mice with B alone, B/αGalCer, B/pep, or B/αGalCer/pep and then determined in vivo CTL activity. As shown in Fig. 2B, only B/αGalCer/pep completely lysed peptide-pulsed targets and, to our surprise, it also maintained complete cytotoxicity even 5 weeks after a single vaccination. Echoing this finding, only the B/αGalCer/pep-treated group showed a significant increase in the number of IFN-γ-producing CD8 T cells against the peptide (Fig. 2C , left).

In another experiment, we additionally primed the B-cell-vaccinated mice with ovalbumin-coated syngenic splenocytes to delineate the CTL responsiveness against subsequent immunization with the same antigen. Mice given B alone or B/αGalCer responded normally toward the priming and generated substantial peptide-specific IFN-γ-producing CD8 T cells (Fig. 2C , right). However, mice given B/pep showed no increase in the number of IFN-γ-producing CD8 T cells, suggesting that these mice were tolerant of the peptide. By contrast, mice vaccinated with B/αGalCer/pep displayed far greater peptide-reactive CD8 T cells than did either the group receiving B/αGalCer or B alone, suggesting that this is a recall response. Based on these collective findings, we concluded that the loading of αGalCer on B/pep generated long-lasting memory cytotoxic immunity.

B/αGalCer/pep is as efficient a generator of CTL as DC/αGalCer/pep or DC/pep. We next sought to compare the efficacy of our B-cell-based vaccine strategy at generating cytotoxicity with that of DC-based vaccine. To this end, we determined the minimum cell number required for achieving complete target lysis in vivo. Serial dilutions of B/αGalCer/pep or DC/αGalCer/pep were i.v. injected into syngenic mice, and an in vivo CTL assay was done. As depicted in Fig. 3A, mice injected with DC/αGalCer/pep showed complete target cell lysis with as few as 16,000 cells. Of interest, a single vaccination with 80,000 B/αGalCer/pep cells was enough to establish a complete peptide-specific lysis, whereas vaccination with 16,000 cells generated a moderate cytotoxicity. However, given that the surface area of DCs is far larger than that of B cells, B/αGalCer/pep may be as efficient as DC/αGalCer/pep in generating cytotoxicity.

DC/pep efficiently generated ovalbumin-specific cytotoxicity, whereas B/pep treatment, regardless of the number of cells tested, did not (Fig. 3B). Of note, the pattern of in vivo cytotoxicity of the DC/pep-treated group was very similar to that of the DC/αGalCer/pep-treated group, indicating that the loading of αGalCer onto DCs did not further enhance the vaccine efficacy of DC/pep in the system.

Generation of CTL by B/αGalCer/pep does not require CD4 T or NK cells but requires CD8 T and iNKT cells. We next examined which types of immune cell were involved in the generation of the CTL response. We injected mice with anti-CD4, anti-CD8, or anti-NK1.1 antibodies 4 days before or 4 days after vaccination with B/αGalCer/pep. Flow cytometric analysis showed that these antibodies efficiently depleted their respective populations in vivo (Supplementary Fig. S2). As it turned out, the timing of depletion made no difference, as the generation of CTL activity was not hampered in either case by the depletion of CD4⁺ or NK1.1⁺ cells (Fig. 4A). It is noteworthy that, although the injection of anti-NK1.1 antibody depleted NK1.1⁺ cells, αGalCer/CD1d-dimer-positive cells were still detectable (Supplementary Fig. S2). As expected, CD8 depletion completely blocked the killing of target cells. Consistent with these results, MHC II^−/− mice (lacking CD4 T cells) developed normal CTL responses, whereas Jα281^−/− mice (lacking iNKT cells) failed to do so (Fig. 4B). In short, the generation of the CTL immunity required both CD8 T and iNKT cells but not CD4 T nor NK cells.

B cells act as real APCs rather than a peptide reservoir, and loading of αGalCer and peptide on the same B cell is required for CTL generation. It could be argued that peptide-pulsed B cells in our model act not as APC but as reservoirs of peptide from which the host DCs withdraw peptides to induce CTL responses. To explore this possibility, we used bm-1 mice model. The cells of these mice can load ovalbumin_257-264 peptide onto their MHC I, but the resulting complex is not recognized by the cognate CD8 T cells due to a mutation in the H-2K region (31). As shown in Fig. 4C, mice vaccinated with B/αGalCer/pep derived from B cells of bm-1 mice failed to develop ovalbumin-specific cytolytic activity, suggesting that DC or other professional APC in the recipient mice was not responsible for the CTL generation in this system.

We next asked if it were possible to generate CTL when αGalCer and peptide were pulsed separately and then injected together. To this end, C57BL/6 mice were i.v. injected with ‘B/αGalCer plus B/pep’ or B/αGalCer/pep alone. As shown in Fig. 4D, mice vaccinated with B/αGalCer plus B/pep failed to generate in vivo cytotoxicity, showing that peptide and αGalCer must be presented on the same B cell to generate the cytotoxicity.

B/αGalCer/pep establishes antitumor immunity. Finally, we asked whether vaccination with B/αGalCer/pep would generate antitumor immunity. To test prophylactic antitumor activity, groups of mice were vaccinated once with B alone, B/αGalCer, B/pep, B/αGalCer/pep, DC/pep, or DC/αGalCer/pep before an ovalbumin-transfected B16 melanoma (MO-5) was transplanted s.c. into them. We observed a slightly delayed pattern of tumor growth in mice vaccinated with B/αGalCer, although all mice finally developed tumors (Fig. 5A). In contrast, no mice receiving B/αGalCer/pep, DC/pep, or DC/αGalCer/pep developed tumor growth. To examine whether these mice established long-term antitumor activity, we rechallenged the surviving mice with s.c. MO-5 tumors s.c. 70 days after the first tumor inoculation. As depicted in Fig. 5B, we observed no tumor growth in those mice, showing that vaccination with B/αGalCer/pep established memory immunity against the tumor.

We next asked whether vaccination with B/αGalCer/pep would eradicate a preexisting tumor. To this end, we established two therapeutic models: mice were vaccinated (a) 1 or (b) 9 days after s.c. transplant when tumors had become palpable. In the 1-day model, vaccination with either DC/pep or DC/αGalCer/pep almost completely suppressed tumor growth (Fig. 5C). Interestingly, tumor growth was also completely diminished in mice vaccinated with B/αGalCer/pep. In the 9-day model, none of these vaccinations completely destroyed the growing tumor due to the aggressive nature of the B16 melanoma. However, in mice vaccinated with B/αGalCer/pep, tumor growth was less pronounced than in ‘B alone’ group of mice and resembled that observed in the DC/pep-vaccinated or DC/αGalCer/pep-vaccinated groups (Fig. 5D).

To determine whether this B-cell-based vaccine regimen can be applied to real tumor antigen, we chose the HER-2/neu model because this tumor antigen is well characterized and its CTL epitope is known (32). Again, we observed a significant level of HER-2/neu-specific cytotoxicity in vivo in mice given αGalCer-loaded HER-2/neu_63-71-pulsed B cells (Fig. 6A). To examine antitumor activity in this model, we injected HER-2/neu-expressing colon carcinoma (CT26-HER-2/neu; ref. 29) i.v. or s.c. into BALB/c mice before vaccinating them with HER-2/neu_63-71-pulsed B/αGalCer. After i.v. tumor inoculation, survival rates were slightly better for those mice vaccinated with B/αGalCer or B/pep than those vaccinated with B alone (Fig. 6B). In sharp contrast, all mice vaccinated with B/αGalCer/pep survived the duration of the experiment. We observed very similar results in the s.c. tumor growth model (Fig. 6C). Collectively, our B-cell-based vaccine regimen proved to be as effective as DC-based vaccines in generating both prophylactic and therapeutic antitumor immunity.

Although several studies have shown that iNKT cell–based vaccines induce antitumor immunity through DC-dependent mechanism (21, 22, 25, 33), we attempted to test whether B cells could be used as a source of cellular vaccine for generating antitumor T-cell immunity in this study. Our findings suggest that iNKT cells can convert tolerogenic B cells into immunogenic APCs that can generate long-lasting cytotoxic immunity. To our surprise, the CTL responses induced by our B-cell-based vaccine were comparable with those induced by DC-based vaccine in efficiency as well as prevention and suppression of tumor growth in several different tumor models. These findings suggest that B-cell-based cellular vaccines would provide an additional resource in the fight against tumors and infectious pathogens.

B cells are known as tolerogenic APCs for CD8 T cells (7). Furthermore, a recent study has shown that B cells suppress iNKT activation when αGalCer is given as free form (34), suggesting that B cells are also poorly immunogenic for iNKT cells. In the current study, however, we observed that cell-associated form of αGalCer on B cells directly activates iNKT cells in vivo as well as in vitro. This discrepancy in findings could be attributed to the different forms of αGalCer (free form versus cell-associated form). It is also possible that the density of αGalCer on B cells may affect the level of iNKT activation. B cells are activated when iNKT cells recognize the αGalCer they harbor. These ‘licensed’ B cells, probably together with other cytokines from iNKT cells, are then able to fully activate cytotoxic T cells whose TCRs are specific for the peptide presented on the surface of the same B cells.

Essential to the design of any vaccine is an antigen-specific long-term memory response. Strikingly, our B-cell-based vaccine approach triggered long-lasting memory cytotoxic immunity, whereby vaccinated mice successfully resisted rechallenge with the same tumor long after the initial tumor had been eradicated. It is generally accepted that CD4 T help is required to elicit efficient CD8 T-cell recall response. However, in the study reported here, primary CTL responses did not require CD4 T cells. Based on these findings, we tentatively propose that iNKT cells acted in place of CD4 T help producing and triggering enough signals to generate memory CTL responses.

It is important to note that cytotoxic activity in this study did not depend on NK1.1⁺ cells. However, we believe that NK cells weakly contributed to the suppression of tumor growth in our tumor challenge model because B cell pulsed with αGalCer alone delayed tumor growth (35). NK1.1 depletion does not mean that the entire iNKT population has been eradicated; despite their name, a segment of the iNKT population does not express NK1.1, and NK1.1⁺ iNKT cells down-regulate NK1.1 after activation (36–38). Thus, the activation of the NK1.1⁻ iNKT population alone would be enough to generate equally effective CTL responses in vivo in NK-depleted mice.

Although several groups have investigated the use of CD40 agonists as adjuvants (4, 6, 10, 11), we instead tested αGalCer because (a) αGalCer activates B cells to express costimulatory molecules with the help of iNKT cells (39), (b) in our previous study, αGalCer but not CD40 agonists were shown to reverse the induction of peripheral tolerance (27, 28, 40, 41), and (c) iNKT can complement the costimulation (signal 2) provided by activated B cells by promptly producing various cytokines and chemokines (signal 3) that enhance T-cell immunity (25). When we used anti-CD40 antibody–activated, peptide-loaded B cells, we failed to observe impressive CTL activity in our system (Supplementary Fig. S3). Interestingly, we observed an impressive up-regulation of CD86 but not CD80. A recent study has shown that CD86 preferentially interacts with CD28, whereas CD80 shows a preference for CTLA-4 (42). Thus, the early expression of CD86 without CD80 on peptide-presenting B cells could contribute to the successful induction of CTL in the current model. Further studies are needed to elucidate which factors are involved in generating CTL by this B-cell-based vaccine.

Theoretically, however, DCs have unique advantages as a source for cellular vaccine because they can capture and process both particulate and soluble antigens and cross-present peptides from those antigens efficiently. Thus, our ‘B-cell-based vaccine’ approach would be useful for the defined CTL epitopes. In fact, several CD8 T-cell epitopes for tumor-associated antigen have been identified and used in clinical trials (43). Moreover, innovative techniques, such as protein transduction by TAT-fusion protein and antigen-expressing viral vector, could offset and compensate for the less efficient capture and cross-presentation of antigen by B cells, allowing them to act as CTL-stimulating APCs against antigens whose CTL epitopes are not defined (11, 12, 44). Although B cells are not a major population in blood, they can be increased exponentially ex vivo by CD40 ligation and maintained for a long period (4–6), obvious clinical assets.

To our knowledge, the current study is the first to show that, with the help of iNKT cells, B cells can induce cytotoxic immunity as well as preventive and therapeutic antitumor immunity as effectively as DCs. Because they also express an abundance of MHC II, B cell could plausibly be made to induce effector CD4 T cells using the procedure outlined above. Recent studies have shown that tumors of blood origin, such as B-cell malignancy and myeloid leukemia, expressed functional CD1d (45, 46). It will be interesting to address whether loading of NKT ligand on these tumors would lead to cytotoxic immunity against their respective natural tumor antigens. For these reasons, we believe that a cellular vaccine strategy using B cell as APC would offer a promising tool for immunotherapy against tumors and infectious pathogens.

Grant support: National R&D Program for Cancer Control, Ministry of Health and Welfare, Republic of Korea grant 0420090-1 (C-Y. Kang).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Albert Bendelac (University of Chicago, Chicago, IL) for DN32.D3 hybridoma, Dr. Se-Ho Park for MHC II–deficient mice, Dr. Doo-Hyun Chung for Jα281-deficient mice, and Dr. Chen Dong (M.D. Anderson Cancer Center, Houston, TX) for critical review and discussion of this article.