Immune Regulation through Mitochondrion-Dependent Dendritic Cell Death Induced by T Regulatory Cells

During T cell development in the thymus, T cells that recognize self-MHC molecules presenting autoantigens can survive by positive selection, whereas highly self-reactive T cells are subsequently deleted by negative selection (1–4). However, T cells that undergo successful maturation and eventually populate the peripheral lymphoid organs carry a certain degree of autoreactivity. It is essential to keep these mature T cells in check to maintain peripheral tolerance. Programmed cell death of mature lymphocytes is a major mechanism for the maintenance of lymphocyte homeostasis and peripheral tolerance (5–7). In addition, natural regulatory T cells (Tregs) that express Foxp3 were established to play an essential role in the protection of immune tolerance (8–14).

Dendritic cells (DCs), the most efficient APCs, are important regulators of both innate and adaptive immune responses (15–19). DCs may also play important roles in the maintenance of immune tolerance (20, 21). We previously observed that DC-specific expression of the baculoviral caspase inhibitor, p35, led to inhibition of Fas-mediated apoptosis in DCs and the development of systemic autoimmune symptoms (22). Consistently, knockout of Fas in DCs also induced the onset of autoimmunity in mice (23). Interestingly, interactions of Ag-pulsed DCs with the Ag-specific T cells may lead to accelerated loss of DCs in vivo (24). It is possible that Fas-dependent killing of DCs by activated T cells provides a negative-feedback mechanism that helps to terminate the activation of lymphocytes by Ag-bearing DCs (25).

It was recognized that DCs have a short lifespan since the original discovery of this cell type (26). The short half-life of DCs in vivo has been linked to an active mitochondrion-dependent cell death pathway regulated by Bcl-2 family members (27–29). The Bcl-2 family members are upstream regulators of the mitochondrion-dependent apoptosis pathway (30, 31). They share the Bcl-2 homology (BH) domains and are divided into three subfamilies (30, 31), including proapoptotic Bax and Bak; the antiapoptotic subfamily proteins, such as Bcl-2, Bcl-xL and Mcl-1; and the proapoptotic BH3-only subfamily, such as Bim, Bid, and Bad. BH3-only proteins are the upstream sensors for different apoptosis signaling in specific cell types (32). BH3-only proteins initiate mitochondrion-dependent apoptosis by either inhibiting the antiapoptotic molecules or directly activating proapoptotic Bax or Bak to induce apoptosis (32, 33). Although deficiency in either Bax or Bak has no salient phenotypes, knockout of both Bax and Bak abolishes mitochondrion-dependent apoptosis (34), indicating that Bax and Bak are essential for mitochondrion-dependent apoptosis but functionally redundant of each other. Both negative selection of developing T cells in the thymus and apoptosis of mature T cells are defective in Bax^−/−Bak^−/− mice (35), suggesting that Bax- and Bak-dependent mitochondrial apoptosis in T cells is involved in the regulation of lymphocyte homeostasis and immune tolerance.

The rapid turnover rates of DCs in vivo could be attributed to the active mitochondrion-dependent cell death pathway in DCs (27). Indeed, transgenic expression of antiapoptotic Bcl-2 or deletion of proapoptotic Bim can inhibit spontaneous cell death in DCs (28, 29). In contrast, Fas-mediated signaling is not required for spontaneous cell death in DCs (29). Therefore, it is possible that mitochondrion-dependent and Fas-mediated cell death pathways regulate DC turnover at different phases of immune responses. However, the function of the active mitochondrion-dependent cell death pathway of DCs in immune regulation has not been well characterized. Furthermore, it is not known whether mitochondrion-dependent cell death in DCs plays a role in the regulation of immune tolerance.

Tregs can interact with DCs to inhibit the activation of Ag-specific T cells in vivo (36–42). Tregs may downregulate important costimulatory molecules on DCs (40, 43–48). Interestingly, one report suggested that Tregs may cause the disappearance of DCs in the draining lymph nodes (38). It has not been determined whether Tregs can induce cell death in DCs or whether such interactions help to protect immune tolerance. Using mice with DCs deficient in Bax and Bak, we showed that CD4⁺Foxp3⁺ Tregs efficiently induced mitochondrion-dependent cell death in DCs. Different from the killing of DCs by effector T cells (Teffs) through Fas, our current study suggests that Tregs exploit the active mitochondrion-dependent apoptosis pathway in DCs for immune regulation. Such interactions between Tregs and DCs potentially play a fundamental role in the regulation of initiation and expansion of Ag-specific immune responses, as well as in the protection of immune tolerance.

Bak^−/−, Bax^flox, Bim^−/−, perforin^−/−, lpr, CD11c-cre transgenic, and OVA-specific–OT1 or -OT2 transgenic mice were obtained from the Jackson Laboratory and maintained on the C57BL/6 background. Granzyme A^−/−granzyme B^−/− mice and wild-type (WT) controls on the 129 × 1/SvJ background were also obtained from the Jackson Laboratory. Foxp3^GFP knock-in mice were provided by Dr. Alexander Rudensky (Memorial Sloan-Kettering Cancer Center, New York, NY) (49). Foxp3^GFP mice were crossed with perforin^−/− mice to generate perforin^−/−Foxp3^GFP mice, as well as with OT2 mice to generate OT2-Foxp3^GFP mice. OT2 mice were also crossed with CD45.1 congenic mice. The mice were maintained in a specific pathogen-free facility and used with the approval of the Institutional Animal Care and Use Committee at Baylor College of Medicine.

Flow cytometry analyses of different cell types, preparation of bone marrow-derived DCs (BMDCs) and splenic DCs, histochemistry, measurements of spontaneous cell death in DCs, DC turnover in vivo by BrdU labeling, and detection of autoantibodies in mice were performed as described (22, 29).

Mice were immunized with OVA (50 μg/mouse) emulsified in CFA at the footpad. Ten days later, total cells (2 × 10⁵/well) from the draining popliteal lymph nodes were cultured in 96-well plates with various concentrations of OVA for 72 h. The cells were pulsed with 1 μCi/well [³H]thymidine for the last 12 h and harvested to measure [³H]thymidine incorporation.

CD4⁺Foxp3⁺ Tregs were sorted from Foxp3^GFP mice and expanded in vitro with anti-CD3– and anti-CD28–coated Dynabeads (Invitrogen; 10 μl beads/10⁶ cells) in the presence of 1000 U/ml IL-2 for 3 d. WT or Bax^−/−Bak^−/− DCs were either unpulsed or pulsed with OVA_323–339 peptide. DCs (10³/well) were mixed with freshly sorted CD4⁺CD25⁻ T cells from OT2 mice (10⁴/well) in 96-well U-bottom plates, in the presence of different numbers of expanded Tregs. The proliferation of OT2 T cells was measured by [³H]thymidine incorporation 4 d later. Alternatively, CD4⁺CD25⁻ Teffs were sorted from WT mice by flow cytometry. Different numbers of Tregs, expanded as above, were added to CD4⁺CD25⁻ Teffs (5 × 10⁴/well) at different ratios in the presence of 0.25 μg/ml soluble anti-CD3 (2C11; BD Biosciences) and irradiated (3000 rad) T cell- and DC-depleted syngeneic splenocytes (5 × 10⁴/well) in 96-well U-bottom plates. [³H]thymidine incorporation was measured 3 d later.

CD4⁺ T cells from OT2 mice (CD45.1 × CD45.2) were sorted and labeled with 5 μM CFSE (Invitrogen) at room temperature for 10 min. The CFSE-labeled OT2 T cells were injected into C57BL/6 mice retro-orbitally (2 × 10⁶/mouse). DCs (2 × 10⁵), pulsed or not with OVA_323–339 peptide, were injected intradermally at the footpad 24 h later, with or without Tregs. Four days later, popliteal lymph nodes were collected, and CD4⁺CD45.1⁺ T cells were analyzed by flow cytometry to determine CFSE dilution.

CD11c⁺I-Ab⁺ DCs were sorted from the spleens of DC-double knockouts of Bax and Bak (DKO) or control mice. DCs were cultured or not in RPMI 1640 complete medium for 24 h. Spontaneous cell death of DCs was quantitated, as described (29). BMDCs or CD11c⁺CD11b⁺ splenic DCs from WT, DC-DKO, or Bim^−/− mice were labeled with 1 μM CFSE (Invitrogen) for 10 min at 37°C. The killing of DCs was analyzed similarly to the previously described protocol (50). Tregs were expanded as above. In some experiments, CD4⁺CD25^high Tregs were sorted from granzyme A^−/−granzyme B^−/− mice or WT controls and expanded as above. Tregs were mixed with DCs (2 × 10⁴/well) at different ratios in 96-well U-bottom plates for 6 h. The cells were incubated with 5 ng/ml 7-amino-actinomycin D (7-AAD; BD Biosciences) at room temperature for 10 min, followed by flow cytometry. Induction of cell death in DCs was quantified essentially as described (50), with the following formula: percentages of killing of DCs by T cells = 100% × (DC_control − DC_T)/DC_control, with DC_control and DC_T representing CFSE⁺7-AAD⁻ DCs in the absence or presence of T cells, respectively. In some experiments, rat IgG or 10 μg/ml blocking Ab to lymphocyte-activation gene 3 (LAG3) (clone C9B7W; BD Biosciences), LFA-1 (clone M17/4; BD Biosciences), or I-A/I-E (clone M5/114.15.2; BioLegend) was added to the culture.

To measure mitochondrial membrane potential (ΔΨm), DCs, with or without incubation with Tregs above, were incubated with 10 μM tetramethylrhodamine ethyl ester (TMRE; Invitrogen) at 37°C for 20 min, followed by flow cytometry. In some experiments, DCs were incubated with 10 μg/ml agonist anti–I-A/I-E (clone 2G9; BD Bioscience) or control rat IgG for 6 h, followed by TMRE staining and flow cytometry.

Splenic CD11c⁺ DCs were purified with anti-CD11c MACS beads (Miltenyi Biotec). Activated CD4⁺ T cells were generated by stimulating sorted CD4⁺CD25⁻ cells anti-CD3– and anti-CD28–coated Dynabeads in the presence of 100 U/ml IL-2 for 3 d. B cells were purified with anti-CD19 MACS beads and stimulated with 1 μg/ml LPS and 100 μM CpG for 48 h. Treg-mediated killing of splenic DCs and activated T or B cells were measured as the killing of BMDCs above.

CD4⁺Foxp3⁺ Tregs or CD4⁺Foxp3⁻ Teffs sorted from OT2-Foxp3^GFP mice were stimulated with OVA_323–339 peptide-pulsed DCs (T/DC = 2:1) for 3 d. OT2 Tregs or Teffs were then isolated by removing DCs with anti-CD11c MACS beads (Miltenyi Biotec) and used for killing of DCs, with or without pulsing with OVA_323–339 peptide. Percentages of killing of DCs by T cells were measured, as above.

To detect interactions between Tregs and DCs in vivo, DCs and Tregs were labeled at 37°C for 10 min with 10 μM 5-(and -6)-(((4-chloromethyl) benzoyl) amino) tetramethylrhodamine (CMTMR) and CFSE (Invitrogen), respectively. WT or DKO DCs (1 × 10⁶) were injected into the footpad of recipient mice. Tregs (0.5 × 10⁶) were also injected into the footpad of some recipient mice. Draining (popliteal) lymph nodes were harvested 24 h later, and frozen sections were analyzed using an LSM 510 confocal microscope (Zeiss).

To determine the conjugate formation between DCs and Tregs in vitro, DCs and Tregs were labeled at 37°C for 10 min with 0.3 μM CMTMR and CFSE, respectively. The cells were mixed at a ratio of 1:1 and centrifuged at 500 × g for 5 min. The cells were incubated at 37°C for 0, 1, 2, or 3 h in the absence or presence of 10 μg/ml anti-LAG3, anti–LFA-1 (BD Bioscience), or rat IgG. The cells were washed with PBS and analyzed by flow cytometry.

To determine autoantibody production after adoptive transfer, WT or Bax^−/−Bak^−/− BMDCs (10⁶ or 5 × 10⁶/mouse) were injected i.p. into 8-wk-old C57BL/6 mice (six mice/group), with or without Tregs (with Treg/DC at 0.5:1), essentially as described (29, 51). The mice were then injected with LPS (30 μg/mouse) i.p. 1 d later. Sera were collected from the recipient mice 1 wk after DC transfer. Anti-nuclear Abs (ANAs), anti-ssDNA, and anti-dsDNA were measured, as above. In parallel experiments, CFSE-labeled WT or Bax^−/−Bak^−/− BMDCs (10⁶ or 5 × 10⁶/mouse) were injected into 8-wk-old C57BL/6 mice (six mice/group), with or without Tregs (with Treg/DC at 0.5:1), at the footpad. Twenty-four hours later, draining (popliteal) lymph nodes were collected. Total cell numbers were counted, and percentages of CFSE⁺ DCs were determined by flow cytometry. Total CFSE⁺ DCs in the draining lymph nodes were calculated.

To determine Treg-mediated signaling in DCs, WT or DKO DCs were incubated with Tregs at a ratio of 2:1 for 6 h at 37°C. The cells were incubated with anti-CD11c MACS beads (Miltenyi Biotec) to isolate DCs. DCs were then lysed for Western blot analyses. The following primary Abs were used: polyclonal rabbit Abs to caspase-8, caspase-3, Bcl-xL, Bad (Cell Signaling), Bax (Santa Cruz Biotechnology), Bcl-2, Bak (Upstate Biotechnology), Mcl-1 (Fitzgerald), Bim (Stressgen), Bid (Imgenex), Blk or Bmf (Biovision), or mouse mAb to caspase-9 (MBL), Noxa (Imgenex), or XIAP (BD Bioscience). The blots were then probed with HRP-conjugated secondary Abs and developed using the chemiluminescent method (Pierce). CD11c⁺CD11b⁺ DCs, CD3⁺ T cells, or CD19⁺ B cells sorted from the spleen of DC-DKO and control mice were also used for Western blot analyses of Bax and Bak, as above. The blots were also probed with anti–α-tubulin (Santa Cruz Biotechnology) to ensure equal loading.

CD11c⁺ BMDCs (10⁶/ml) were cultured in the absence or presence of 1 μg/ml LPS (Sigma) or 0.5 μM phosphorothioate-stabilized CpG oligonucleotide (5′-TCCATGACGTTCCTGATGCT-3′) for 24 h. Brefeldin A (1 μg/ml) and monensin (2 μM) were added during the last 6 h to inhibit cytokine secretion. Cells were stained with FITC–anti-CD11c, followed by fixation and permeabilization with Cytofix/Cytoperm solution (BD Biosciences) and staining with PE-conjugated anti–IL-12p40/p70 (BD Biosciences), PE-conjugated IL-6 (BD Biosciences), or PE-conjugated rat IgG1 as isotype control. The cells were then analyzed by flow cytometry.

Data are presented as mean ± SD. The p values were determined by two-tailed Student t test using GraphPad Prism software version 4.0 for Macintosh; p < 0.05 was considered statistically significant.

To determine the functions of mitochondrion-dependent apoptosis in immune regulation, we generated double knockouts of Bax and Bak in DCs (DC-DKO) by crossing CD11c-cre mice with Bax^flox and Bak^−/− mice. Specific deletion of Bax in DCs was observed in DC-DKO mice (Supplemental Fig. 1A). Because Bax and Bak are functionally redundant of each other (34), deletion of Bak alone is not expected to affect apoptosis in other cell types. DC-DKO mice, but not Bak^−/− or DC-Bax^−/− mice, displayed enlargement of the spleen and lymph nodes (Supplemental Fig. 1B). Terminally differentiated DCs have low proliferative potentials; pulsing with BrdU in vivo can be used to measure the rate of DC turnover by the appearance of newly generated BrdU⁺ DCs (52). We found that DCs were labeled more slowly with BrdU in DC-DKO mice (Supplemental Fig. 1C), indicating a reduced DC turnover rate in these mice. Consistently, the percentages and total numbers of CD11c^highI-Ab⁺ or CD11c^highCD40⁺ conventional DCs were increased in DC-DKO mice (Fig. 1A, 1B). CD11c^lowPDCA-1⁺ plasmacytoid DCs increased to a lesser extent in DC-DKO mice (Fig. 1A, 1B). The percentage of T, B, or NK cells was not elevated in DC-DKO mice (Supplemental Fig. 1D). DCs from DC-DKO mice, but not Bak^−/−, DC-Bax^−/−, Fas-deficient lpr, CD11c-cre, or C57BL/6 control mice, showed defects in spontaneous cell death (Fig. 1C). However, splenic DCs from DC-DKO mice and WT controls expressed similar levels of CD40, MHC class II (MHC-II), B7.1, B7.2, and ICAM-1 (Supplemental Fig. 1E), suggesting that deficiencies in Bax and Bak cause DC accumulation but not abnormal DC activation.

T cells from DC-DKO mice showed increased expression of an activation marker, CD69 (Supplemental Fig. 2A). In particular, more than half of CD4⁺ T cells were CD69⁺ (Supplemental Fig. 2A). CD69 was also increased on CD8⁺ T cells and CD19⁺ B cells in DC-DKO mice (Supplemental Fig. 2A). In addition, similar to lpr mice, T cells with the CD44⁺CD62L⁻ activated/memory phenotype were increased in DC-DKO mice compared with controls (Supplemental Fig. 2A). These data suggested that deficiencies in Bax and Bak in DCs lead to DC expansion and abnormal lymphocyte activation.

We determined whether deficiencies in mitochondrion-dependent cell death induce autoimmune responses in DC-DKO mice. DC expansion and enlargement of lymphocyte areas were observed in spleen sections of-6 mo-old DC-DKO mice (Supplemental Fig. 1F). Severe perivascular lymphocyte infiltration was found in the liver, lung, and kidney of DC-DKO mice but not DC-Bax^−/−, Bak^−/− or WT controls (Supplemental Fig. 1G). IgG deposits were also observed in the glomeruli of kidneys in DC-DKO mice (Supplemental Fig. 1H). We also detected autoantibodies, including anti-dsDNA, anti-ssDNA, and ANAs, in the sera of 3- and 6-mo-old DC-DKO mice, but not in controls, by ELISA (Fig. 2A). Consistent with the production of ANAs, sera from DC-DKO mice showed nuclear staining of Hep2 cells (Fig. 2B). Together, these observations suggested that deficiencies in Bax and Bak in DCs lead to the development of systemic autoimmunity. Normal levels of CD4⁺Foxp3⁺ natural Tregs were detected in DC-DKO mice (Supplemental Fig. 2A). Moreover, natural Tregs from DC-DKO and control mice showed comparable activities in inhibiting the proliferation of Teffs (Supplemental Fig. 2B), indicating that there is no intrinsic defect in natural Tregs in DC-DKO mice.

We found that cells from the draining lymph nodes of immunized DC-DKO mice responded better to Ag restimulation than those of WT, Bak^−/−, or DC-Bax^−/− mice (Fig. 3A). Consistently, Bax^−/−Bak^−/− DCs induced more robust proliferation of Ag-specific T cells than did control DCs in vivo, as assayed by CFSE dilution (Fig. 3B). This suggested that Bax^−/−Bak^−/− DCs are more efficient in priming Ag-specific T cells in vivo. However, Bax^−/−Bak^−/− DCs did not produce more IL-6 or IL-12 (Supplemental Fig. 2F), suggesting that Bax^−/−Bak^−/− DCs do not overproduce these cytokines to induce more activation of Teffs. Interestingly, the increase in the capacity for Bax^−/−Bak^−/− DCs to stimulate Ag-specific Teffs in vitro was detectable but less dramatic (Fig. 3C, 3D, Supplemental Fig. 2D, 2E). This is reminiscent of previous observations that cell death in DCs has a more profound effect on the immunogenicity of DCs in vivo than in vitro (53). Such differences between in vivo and in vitro observations could be due to the presence of other cell types that regulate the interactions between DCs and T cells in vivo.

Although natural Foxp3⁺ Tregs were not reduced or dysfunctional in DC-DKO mice (Supplemental Fig. 2A, 2B), it remains possible that Bax^−/−Bak^−/− DCs might be refractory to suppression by Tregs to induce uncontrolled T cell activation. We investigated whether deficiencies in Bax and Bak might enable DCs to overcome the inhibitory effect of Tregs. We sorted and expanded CD4⁺Foxp3⁺ natural Tregs from Foxp3^GFP knock-in mice (49). To the culture of Ag-pulsed DCs and OVA-specific OT2 Teffs (with DCs:Teff at a fixed ratio of 0.1:1), we added varied numbers of the expanded Tregs (with Treg:Teff ratios ranging from 0.01:1 to 1:1). When Foxp3⁺ Treg numbers were low (Treg:Teff from 0.01:1 to 0.1:1), we detected significant suppression of Ag-specific T cell proliferation induced by WT, Bax^−/−, and Bak^−/− DCs but not Bax^−/−Bak^−/− DCs (Fig. 3E). When Treg:Teff ratios were higher (0.1:1–1:1), significant suppression of Bax^−/−Bak^−/− DC-induced proliferation of OT2 Teffs was detected (Fig. 3E). However, this could be attributed, at least in part, to a direct suppression of Teffs by Tregs at higher Treg/Teff ratios (Supplemental Fig. 2C).

Consistent with in vitro observations (Fig. 3E), CD4⁺Foxp3⁺ Tregs suppressed the proliferation of OT2 Teffs induced by WT, Bax^−/− or Bak^−/− DCs in vivo, as measured by CFSE dilution (Fig. 3F). In contrast, Bax^−/−Bak^−/− DC-induced T cell proliferation was less susceptible to inhibition by Tregs in vivo (Fig. 3F). This supports the conclusion that Bax^−/−Bak^−/− DCs are refractory to suppression by Tregs.

We found that activated Tregs acquired cytotoxicity against DCs (Fig. 4A). We examined whether activated Tregs could directly induce cell death in DCs in a Bax- and Bak-dependent manner. We observed that DCs were more susceptible than activated B or T cells to killing by CD4⁺Foxp3⁺ Tregs (Fig. 4B). WT, but not Bax^−/−Bak^−/−, DCs were sensitive to killing by Tregs (Fig. 4C; Supplemental Fig. 3A, 3B), suggesting that Tregs induce DC cell death in a Bax- and Bak-dependent manner. Treg-mediated killing of DCs is not affected by treating DCs with LPS (Supplemental Fig. 3C), suggesting that Treg-mediated killing of DCs is not affected by DC maturation. It was shown that Tregs can negatively regulate DCs by restricting DC development (54), inhibiting the expression of costimulatory molecules on DCs or competing with Teffs in the interaction with DCs (36–38, 40). Apoptotic DCs may promote further induction of Tregs (55). Inducing Bax- and Bak-dependent cell death in DCs may provide another mechanism for DC regulation by Tregs.

Blocking MHC-II inhibited the killing of DCs by polyclonal Tregs (Fig. 4D). This suggests that the polyclonal Tregs are potentially autoreactive. By using OVA-specific OT2 Foxp3⁺ Tregs, we found that Foxp3⁺ Tregs killed Ag-pulsed, but not unpulsed, DCs (Fig. 4E). Using Foxp3⁺ Tregs expressing a transgenic TCR specific for a foreign Ag may not recuperate the action of natural Tregs carrying TCRs that tend to be autoreactive. Nevertheless, our data obtained by using OT2 Foxp3⁺ Tregs suggested that recognition of Ags on DCs is important for killing by Tregs. In contrast to Tregs, polyclonal Teffs did not efficiently kill DCs in the absence of anti-CD3 (Fig. 4C), whereas OT2 Teffs showed killing of DCs in the presence of OVA Ag at a higher T:DC ratio (3:1; Fig. 4E).

We consistently detected more Bax^−/−Bak^−/− DCs in the draining lymph nodes after adoptive transfer (Fig. 5A). Moreover, cotransfer of Tregs led to the loss of WT, but not Bax^−/−Bak^−/−, DCs in the draining lymph nodes (Fig. 5A). This is consistent with the conclusion that Bax and Bak regulate spontaneous cell death of DCs, as well as Treg-dependent DC killing. Tregs can induce Bax- and Bak-dependent cell death in DCs. In addition, we also observed that Tregs formed conjugates with both WT and Bax^−/−Bak^−/− DCs in the draining lymph nodes (Fig. 5A). Conjugate formation between Tregs and DCs was also detected in vitro (Supplemental Fig. 3D–H). WT and DKO DCs were similar in forming conjugates with Tregs after 1 h of incubation in vitro (Supplemental Fig. 3E, 3H). Consistent with confocal microscopy analyses, more Bax^−/−Bak^−/− DCs were found in the draining lymphocytes than WT DCs, as determined by flow cytometry (Fig. 5B). Cotransfer of Tregs also promoted the clearance of WT, but not Bax^−/−Bak^−/−, DCs (Fig. 5B). These data suggested that Bax and Bak do not affect conjugate formation between Tregs and DCs. Rather, Bax and Bak regulate spontaneous cell death of DCs, as well as Treg-mediated clearance of DCs in vivo.

To directly test whether Bax^−/−Bak^−/− DCs can induce autoimmune responses, we performed adoptive transfer of WT and Bax^−/−Bak^−/− DCs, with or without Tregs. It was shown that transfer of excessive activated DCs can trigger the production of autoantibodies in recipient mice (51, 56). Adoptive transfer of Bax^−/−Bak^−/− DCs at a low dose (1 × 10⁶ cells/mouse) induced the production of anti-dsDNA, anti-ssDNA, and ANAs, whereas a higher dose (5 × 10⁶ cells/mouse) triggered more autoantibody production (Fig. 6). Cotransfer of Tregs did not significantly suppress autoantibody production induced by Bax^−/−Bak^−/− DCs (Fig. 6). In contrast, WT DCs induced detectable levels of autoantibody production only at the high dose (5 × 10⁶ cells/mouse), which was efficiently suppressed by cotransfer of Tregs (Fig. 6). These data provided direct evidence to support the conclusion that Bax^−/−Bak^−/− DCs have the propensity for triggering autoantibody production. Moreover, Bax^−/−Bak^−/−, but not WT, DCs are resistant to suppression by Tregs in the induction of autoimmune responses.

We next investigated apoptosis signaling in DCs induced by Tregs. After incubation with Tregs, WT DCs lost ΔΨm, whereas DKO DCs were relatively resistant to the loss of ΔΨm (Fig. 7A). Consistent with the activation of mitochondrion-dependent cell death in WT DCs, Tregs induced the activation of caspase-9 and caspase-3 in WT, but not Bax^−/−Bak^−/−, DCs (Fig. 7B). It was shown that cleavage of Bid into active truncated Bid by caspases can trigger mitochondrial apoptosis (57). Caspase-dependent cleavage may inactivate antiapoptotic Bcl-2 family proteins to promote mitochondrion-dependent cell death (58). However, we did not find processing of Bcl-2 family proteins in DCs after incubation with Tregs (Supplemental Fig. 4A). Tregs also did not change the expression of Bcl-2 family proteins in DCs (Fig. 7B, Supplemental Fig. 4A). Interestingly, Bim^−/− DCs showed resistance to killing by Tregs compared with WT DCs, whereas deletion of either Bax or Bak in DCs did not affect their killing by Tregs (Fig. 7C). This suggested that Bim serves as a mediator to activate Bax and Bak in DCs after encountering Tregs. Although the precise mechanism for the activation of Bim is not resolved, Bim potentially transmits cell death signaling by sequestering antiapoptotic molecules or directly activating Bax and Bak (59, 60).

We then investigated which effector molecules in Tregs could induce cell death in DCs. The killing of DCs by Teffs involves Fas and perforin (22, 61, 62) but does not appear to require Bax and Bak (Supplemental Fig. 4B, 4C), suggesting that Tregs and Teffs use different mechanisms to kill DCs. Interestingly, Tregs in tumor tissues can kill DCs in a perforin/granzyme-dependent manner (41), whereas Treg-induced cell death in Teffs is independent of Fas or perforin (63). We found that Treg-mediated killing of DCs did not involve Fas but required prior activation of Tregs (Fig. 4A, Supplemental Fig. 4D). Also, Tregs did not induce proteolytic activation of caspase-8 in DCs (Fig. 7B), consistent with the possibility that Tregs do not engage Fas on DCs to activate caspase-8. In addition, deficiencies in perforin or granzymes A/B in Tregs did not affect the killing of DCs (Supplemental Fig. 4E), suggesting that these molecules are not required for the killing of DCs by natural Tregs.

We observed that blocking LFA-1 inhibited Treg-mediated killing of DCs (Fig. 7D). The conjugate formation between Tregs and DCs was inhibited by blocking LFA-1 (Supplemental Fig. 4G, 4H), suggesting that cell adhesion promotes the killing of DCs by Tregs. Molecules highly expressed by Tregs were suggested to be critical for Treg functions (11, 40, 64, 65). We observed that inhibition of LAG3 partially suppressed the killing of DCs by Tregs (Fig. 7D), whereas neutralization Abs to soluble factors, including IL-10 and TGF-β, had no effect (data not shown). LAG3 is a potent ligand for MHC-II molecules and can trigger negative signaling in DCs to suppress maturation and immunostimulatory capacity of DCs (66). On unstimulated Tregs, the expression of LAG3 was low (Fig. 7E). Tregs acquired cytotoxicity against DCs after activation (Fig. 4A), whereas LAG3 was also significantly upregulated on activated Tregs (Fig. 7E). Increased LAG3 on activated Tregs may enable Tregs to trigger cell death in DCs by engaging MHC-II. Indeed, we observed that engagement of MHC-II on DCs with an agonist Ab triggered the loss of ΔΨm in WT, but not Bax^−/−Bak^−/−, DCs (Fig. 7F). Engagement of MHC-II or other molecules on DCs by activated Tregs may transmit signals into DCs through BH3-only molecules, such as Bim, to induce the activation of Bax and Bak, leading to mitochondrial disruption and cell death.

DCs harbor an active mitochondrion-dependent apoptosis pathway regulated by Bcl-2 family members (27–29). In this study, we determined the functions of mitochondrion-dependent cell death in DCs in immune regulation using mice with DC-specific knockout of Bax and Bak. Bax and Bak deficiencies in DCs resulted in DC expansion, spontaneous T cell activation, and development of systemic autoimmunity. In addition to regulating spontaneous cell death by Bax and Bak, our data suggested another level of regulation of Treg-induced cell death in DCs through Bax and Bak. Expression of LAG3 may enable Tregs to trigger mitochondrion-dependent cell death in DCs. Adoptive-transfer experiments provide direct evidence to show that Bax- and Bak-deficient DCs have increased propensity for inducing the production of autoantibodies. Moreover, Tregs inhibited WT, but not Bax- and Bak-deficient, DCs in the induction of autoantibodies after adoptive transfer. Our data suggested that the Bax- and Bak-dependent pathway is involved in both spontaneous cell death and Treg-mediated killing of DCs, and both of these mechanisms are important for maintaining DC homeostasis and preventing autoimmunity.

Bax- and Bak-dependent spontaneous cell death in DCs may occur throughout the courses of immune responses. In contrast, only preactivated Tregs expressed higher levels of LAG3 (Fig. 7E) and showed the capacity to kill DCs (Fig. 4A), suggesting that Bax- and Bak-dependent killing of DCs may happen only after Tregs are activated. We and other investigators previously demonstrated that DCs are susceptible to Fas-mediated apoptosis (22, 23). Fas-dependent killing of DCs may take place when activated T cells that express high levels of FasL are present. This potentially provides a negative-feedback mechanism for the suppression of DC-dependent activation of T cells at late stages of immune responses, possibly after significant T cell activation induced by DCs has taken place. It was reported that Tregs can kill autologous CD8⁺ T cells or LPS-induced monocytes through FasL/Fas interactions (67, 68). Other studies showed that the suppressive effect of Tregs is not inhibited by neutralization of FasL or using FasL-deficient Tregs (69, 70). We also observed that the killing of DCs by Tregs was independent of Fas (Supplemental Fig. 4). Such differences in the involvement of FasL/Fas interactions could be due to different target cells used in these studies.

Fas-mediated killing is usually dependent on activated T cells that express high levels of FasL. Spontaneous cell death in DCs through mitochondrion-dependent intrinsic apoptosis pathway is regulated by Bcl-2 family members (28, 71). Different from Fas-dependent cell death, Bax- and Bak-dependent spontaneous cell death in DCs may have a more broad influence at the initiation, expansion, and contraction phases of immune responses. In contrast, killing of DCs by activated Tregs through Bax and Bak may only function at the contraction phase of immune responses.

Recognition of Ags by TCRs on Tregs and LFA-1–dependent Treg-DC conjugate formation may be critical for Tregs to kill DCs. Interestingly, activated Tregs expressed elevated LAG3 and acquired killing activities toward DCs, and blocking of LAG3 partially inhibited the killing of DCs by Tregs (Fig. 7). It was shown that LAG3 can trigger negative signaling to inhibit the maturation and immunostimulatory capacity of DCs (66). Interestingly, cross-linking of MHC-II triggered the loss of ΔΨm in DCs in a Bax- and Bak-dependent manner (Fig. 7H). Our data suggested that LAG3 on activated Tregs can engage MHC-II on DCs to induce mitochondrial disruption and cell-death signaling in DCs.

Certain DC subsets can induce the generation of Tregs in vitro and in vivo (20, 72–77). In contrast, Tregs also restrict the development of DCs in vivo (10, 54). Our data suggested that Tregs can restrict DC expansion through Bax- and Bak-dependent killing, and deficiency in the killing of DCs by Tregs contributes to DC accumulation and induction of autoimmunity in DC-Bax^−/−Bak^−/− mice. Therefore, dynamic interplays between DCs and Tregs are important for the maintenance of a balanced immune system. Induction of mitochondrion-dependent cell death in DCs by Tregs may serve as one important mechanism for immune regulation. Promoting the interactions between Tregs and DCs may provide effective avenues to prevent and treat autoimmune diseases, whereas inhibiting the killing of DCs by Tregs may help to boost immune responses to infections and cancer.

We thank Yaming Liang, Yiqing Zhang, Jie Huang, and Shanshan Bai for technical assistance and David Corry and John Rodgers for discussions.

The authors have no financial conflicts of interest.