In vivo costimulatory role of B7-DC in tuning T helper cell 1 and cytotoxic T lymphocyte responses

To knock out the B7-DC gene and simultaneously introduce a reporter for B7-DC expression, we used a knock-out/knock-in construct that replaced B7-DC with the EGFP gene (Fig. 1a). Animals derived from recombinant embryonic stem cells were backcrossed nine generations onto the BALB/c background. Initially, myeloid and lymphoid populations from the thymus, secondary lymphoid tissues, and liver were analyzed. Overall cell numbers and proportions of standard subsets of T, B, NK, NK T cells, and DC were fairly similar in the thymus and spleen of WT and B7-DC KO mice (Fig. S1). This similarity contrasts with the recently reported B7-H1 KO mice, in which the total number of CD8 T cells in the liver is significantly increased (18). Importantly, the expression pattern of MHC class II as well as other B7 family members, including B7-1 (CD80), B7-2 (CD86), and B7-H1, on DCs was not appreciably affected by the absence of B7-DC (Fig. 1b). Likewise, the cell yield from bone marrow–derived DC (BMDC) cultures was essentially identical in WT and B7-DC KO mice (Fig. 1c).

As described in earlier publications, surface expression of B7-DC is largely restricted to DCs and a subset of macrophages, with essentially no detectable expression on B cells or T cells in any activation state (16). Indeed, purified T cells from B7-DC KO mice proliferate in response to the T cell mitogen, Concanavalin A (ConA), equivalently to WT T cells (Fig. 2a). In contrast, naive OVA-specific CD4 T cells (from the DO.11.10 TCR transgenic line) are stimulated to proliferate somewhat less efficiently than WT DCs in vitro by OVA-pulsed B7-DC KO DCs (Fig. 2b). Given that MHC II, B7-1, B7-2, and B7-H1 are not altered on B7-DC KO DCs (Fig. 1b), these data are compatible with a costimulatory role for B7-DC. However, they could also reflect effects on other costimulatory molecules or backward signaling into DCs by B7-DC.

Earlier studies have shown that when plate-bound B7-DC-Fc plays a costimulatory role, particularly when used in conjunction with B7-1-Fc or B7-2-Fc, there is a pronounced effect on IL-2 and IFN-γ, (14). We therefore further evaluated the role of B7-DC in Th1 versus Th2 differentiation upon culturing purified DO11.10 CD4 T cells with BMDCs from WT or B7-DC KO mice under Th1- or Th2-generating conditions in the presence of OVA_323-339 peptide. Fig. 3 shows that T cells cultured with BMDCs from B7-DC KO mice generated roughly 40% fewer IFN-γ⁺ cells than did T cells cultured with BMDCs from WT mice (59.7% vs. 35.0%, respectively). The same reduction in IFN-γ⁺ CD4 T cells was observed in the presence of additional anti-mouse CD3 antibody during restimulation (81.7% vs. 50.6%, respectively). In contrast, culture under Th2-generating conditions failed to demonstrate a significant difference in IL-4⁺ T cells whether WT or B7-DC KO BMDCs were used.

These results support the notion that B7-DC selectively costimulates Th1 responses. The effect is certainly not all-or-none, suggesting the this particular B7 family member tunes IFN-γ responses rather than being absolutely necessary for their generation.

To analyze the role of B7-DC in the priming of in vivo immune responses, we adoptively transferred naive OVA-specific DO11.10 cells and measured responses to immunization with peptide loaded onto DCs. 5 million purified carboxyfluorescein succinimidyl (CFSE)-labeled DO11.10 CD4 T cells were transferred into WT or B7-DC KO mice on d −1. Because it remains unresolved whether priming by ex vivo–loaded DCs involves direct presentation or transfer of antigen from injected DCs to endogenous DCs, the genotype of the BMDCs was matched to the immunized host; that is, WT mice were immunized with WT DCs, and B7-DC KO mice were immunized with B7-DC KO BMDCs. This design assures that all potential DCs presenting antigen in vivo are uniformly WT or B7-DC KO. When these mice were immunized by OVA_323-339 peptide–pulsed BMDCs, a significant effect of B7-DC on T cell expansion was observed. Whereas OVA-specific KJ1-26⁺ T cells expanded to 11.6%, 1.8%, 4.1%, and 7.6% of total CD4 T cells in spleen, peripheral lymph nodes (PLN), liver, and hepatic lymph node (HLN), respectively, expansion in B7-DC KO mice was 7.7% (34% reduction), 0.9% (50% reduction), 2.4% (42% reduction), and 3.0% (61% reduction) in the respective compartments by d 3 after immunization (Fig. 4a). By d 5, B7-DC KO mice showed similar reductions relative to those in WT mice in all compartments except PLN, in which there was no significant reduction (Fig. 4b). A more pronounced effect of B7-DC deletion was observed in the analysis of CFSE dilution in DO11.10 CD4 T cells (Figs. 4, c and d, and Fig. S2). For example, the ratios of undivided cells (no CFSE dilution) to undivided cells were significantly different, particularly in the liver and HLN, on d 3 and d 5. On d 3, there were 8.7 times more undivided cells in B7-DC KO mouse liver than in WT mouse liver and 2.7 times more undivided cells in B7-DC KO mouse HLN than in WT mouse HLN. On d 5 there were 2.9 times more undivided cells in B7-DC KO mouse liver than in WT mouse liver and 2.4 times more undivided cells in B7-DC KO mouse HLN than in WT mouse HLN). Analysis of CD4⁺ KJ1-26⁺ T cell expansion and CFSE dilution showed similar trends in differences between B7-DC KO and WT mice among the various compartments; namely, liver > HLN > spleen > PLN.

To follow up on the in vitro results suggesting that B7-DC selectively contributes to Th1 responses, we analyzed in vivo cytokine responses to immunization with OVA peptide-pulsed DCs in animals receiving DO11.10 CD4 T cells 1 d before immunization, as described previously. To avoid effects on cytokine patterns caused by in vitro stimulation, we analyzed cytokine mRNA by quantitative RT-PCR from OVA-specific T cells purified directly from immunized animals. 5 d after immunization, CD4⁺KJ1-26⁺ T cells were sorted out, and the RNA was prepared for quantitative RT-PCR. Cytokine responses as assessed by mRNA were highly antigen specific—virtually none were detected in purified CD4⁺KJ1-26⁺ T cells after immunization with unpulsed DCs. Significantly less IFN-γ, slightly less IL-2, and significantly more IL-4 was consistently observed in B7-DC KO mice immunized with peptide-pulsed B7-DC KO BMDCs than in WT mice immunized with WT BMDCs (Fig. 5). These results demonstrate that, in the absence of B7-DC, cytokine production is skewed from Th1 to Th2 with decreased IFN-γ and increased IL-4 production. As in the in vitro studies, the effect is partial, compatible with a tuning function for B7-DC.

The patterns of cytokine expression from both the in vitro and in vivo studies described previously suggested that B7-DC has a positive costimulatory role in enhancing Th1-type responses. It was therefore of interest to analyze the effect of B7-DC deletion on antibody isotypes in response to immunization, because isotype switching is dependent on production of specific cytokines by Th cells and, as such, partially reflects Th1/2 balance. WT and B7-DC KO mice were immunized with the T cell–dependent antigen NP¹⁷-KLH in alum, and the serum levels of NP-specific antibody isotypes were measured. Significantly, all classes of antibody were essentially identical in WT and B7-DC KO mice with the exception of the IFN-γ–dependent IgG2a class, which was dramatically reduced in B7-DC KO mice (Fig. 6). These results support a functional role for B7-DC in costimulating IFN-γ–dependent class switching in vivo.

To determine whether B7-DC plays a role in MHC class I–restricted CTL generation, we transferred 2 × 10⁶ T cells from H-2Kd⁺ hemagglutinin (HA)-specific TCR transgenic mice (24, 25) into BALB/c WT or B7-DC KO mice and immunized them with unloaded BMDCs as control or with BMDCs loaded with the cognate peptide (HA_518-526). As in the in vivo CD4 T cell experiments, WT mice were immunized with WT BMDCs, and B7-DC KO mice were immunized with B7-DC KO BMDCs. As with antigen-specific CD4 T cells, expansion of HA-specific CD8 T cells was lower in B7-DC KO mice than in WT mice, particularly in the liver and, to a lesser extent, in the spleen (Fig. 7). As with CD4 T cells, the smallest difference in absolute numbers of HA-specific CD8 T cells in WT and B7-DC KO mice was observed in PLN. To measure CTL activity directly, an in vivo CTL assay was done in which two populations of BALB/c splenocytes differentially labeled with CFSE, one of which loaded with HA_518-526, were transferred into the immunized mice. As a control, some immunized mice received splenocytes in which neither the CFSE^high or CFSE^low cells were peptide loaded. In vivo CTL activity is determined based on the selective reduction in ratio of peptide-loaded cells (CFSE^low) to unloaded cells (CFSE^high) reisolated from the spleen 10 h after injection. As seen in Fig. 7b, in vivo CTL activity in immunized B7-DC KO mice was roughly half that of WT mice. Thus, B7-DC seems to play a costimulatory role in vivo for CTL proliferation and activation, either indirectly by enhancing Th function or directly by costimulating CD8 T cells. Because only the CTL epitope of HA was used in the immunizations, a direct role in CD8 T cell costimulation independent of CD4 T cells is the most likely mechanism. This notion is supported by findings that B7-DC can strongly costimulate purified CD8 T cells in vitro (15).

IFN-γ responses are crucial for both natural and induced antitumor immunity. Tumors arise more frequently and grow more quickly in mice with deletions in either the IFN-γ gene or components of the IFN-γ signaling pathway. In general, strong Th1 responses tend to be linked to strong CD8 CTL responses, and the combination of strong Th1 and strong CD8 CTL responses is typically associated with successful antitumor immunity. We therefore analyzed the growth of hepatic metastases of the well-characterized CT26 colon tumor in WT and B7-DC KO mice as well as the CD8 T cells specific for the immunodominant H-2L^d-restricted CT26 antigen, AH1. CT26 is a partially immunogenic tumor, because injection of WT tumor cells into BALB/c mice elicits an AH1-specific T cell response. This CD8 T cell response slows tumor growth, but the tumor ultimately outgrows the antitumor immune response and kills the host animal. The mortality rate in CT26-bearing mice is roughly proportional to the strength of the natural antitumor response generated, because CD8-depleted animals succumb to tumor faster than WT animals. Fig. 8a demonstrates that B7-DC KO mice bearing hepatic metastases of CT26 die more quickly than WT mice. This mortality is associated with more rapid growth of CT26 tumors in the liver (Fig. 8b). Tetramer staining of AH1-specific CD8 T cells reveals a marked decrease in the number of AH1-specific CD8 T cells in the livers of tumor-bearing B7-DC KO mice relative to WT mice (Fig. 8, c and d). Thus, B7-DC seems to contribute in a positive fashion to the endogenous generation of intrahepatic CT26-specific CD8 responses in animals bearing hepatic CT26 metastases.

We present in vitro and in vivo experiments in B7-DC KO mice that support a predominantly costimulatory role for B7-DC in the generation of both Th1 and CTL responses. In some cases, decreased IFN-γ responses in B7-DC KO mice are associated with increased IL-4 (Th2) responses (Fig. 5), whereas in other cases, such as isotype switching, IL-4–dependent antibody isotypes (IgG1) are unaffected. In all cases, the effects are partial rather than complete. We therefore propose that, unlike a molecule such as T-bet, on which Th1 responses depend absolutely (26), the role of B7-DC is to amplify or tune Th1 responses. Fine-tuning of immune responses seems to be critical so that they do not undershoot or overshoot. The importance of not undershooting the response to an infectious organism is obvious. Conversely, the toxicities of overzealous immune responses are readily apparent in mice deficient in inhibitory checkpoint molecules and in humans treated with antibodies such as anti-CTLA-4 (27).

The most straightforward interpretation of these results is that decreased IFN-γ and CTL responses in B7-DC KO mice are caused by loss of a costimulatory signal to T cells that interacts with B7-1/2 costimulation. However, an additional possibility is that DC activation caused by backward signaling by B7-DC is diminished. The fact that B7-1, B7-2, and surface MHC II are not appreciably different in B7-DC KO mice and WT mice suggests that DC activation status is not appreciably affected by the absence of B7-DC. Nonetheless, a definitive distinction between effects on forward versus backward signaling will require further experimentation.

Taken together, our results strongly refute the notion that B7-DC is a redundant inhibitory ligand for B7-H1, even though both bind PD-1. In particular, the finding of diminished CD4 and CD8 responses in the livers of B7-DC KO mice stands in striking contrast to the increased numbers and responses of activated CD8 T cells in the livers of B7-H1 KO mice (18). Direct comparisons between these two KO models in identical systems and strain backgrounds remain to be made. However, a preliminary picture of distinct roles for these B7 family members seems to be emerging from the different expression patterns seen in B7-DC (predominantly DCs) and B7-H1 (expression on both hematopoietic and nonhematopoietic cells of peripheral organs as well as many tumors) mice and the respective KO mice. Thus, B7-DC seems to have a primarily costimulatory role in immune priming, whereas B7-H1 has a predominantly down-modulating role for activated T cells in peripheral organs and tumors. Further support for this distinct physiology comes from the findings of Liu and colleagues that B7-DC–transfected tumors enhance antitumor immunity in contrast to B7-H1–expressing tumors that are protected from elimination by activated tumor-specific T cells (17, 28). The finding that tumors grow faster and elicitation of tumor-specific CD8 T cells is impaired in B7-DC KO mice further accentuates the physiologic differences between these two molecules and suggests that B7-DC signals may significantly enhance antitumor immunity.

Many questions remain to be answered about B7-DC physiology. Foremost is the identification of its cognate costimulatory receptor. We hypothesize that the costimulatory role of B7-DC observed in vivo is largely mediated by this costimulatory receptor rather than by PD-1. Consistent with this notion, treatment of mice with anti-PD1 antibodies did not alter B7-DC costimulatory effects (unpublished data). However, it is possible that, under certain circumstances, PD-1 may transmit costimulatory rather than inhibitory signals. In addition, it will be important to determine how B7-DC signals interact with other signaling pathways that affect the Th1/Th2 balance. Along these lines, we are crossing the B7-DC KO mice onto a C57/BL6 background, which is more “Th1 prone” than the BALB/c background on which the current studies were performed. A recent study suggested that the IFN-γ receptor differentially localizes into the immune synapse in these two strains (29). Comparison between these two strains may allow us to determine whether the Th1 skewing by B7-DC is affected by interactions with other background genes.

6–8-week-old BALB/c mice and antigen-specific TCR transgenic DO11.10 mice (H-2^d) were purchased from the National Cancer Institute and Jackson ImmunoResearch Laboratories, respectively. Thy1.1 background HA (clone 4) transgenic mice were the gift from E. Fuchs (Johns Hopkins University). B7-DC knock-out/knock-in mice were generated using standard methodologies (30). A genomic fragment containing exon 2 of the B7-DC gene, which includes first methionine, was cloned from a 129/SvJ mouse genomic library. The short and long fragments, two of loxP sites and one of EGFP site (CLONTECH Laboratories, Inc.) were incorporated into pNTKV1901 (Stratagene; Fig.1 a). This targeting vector was linearized and electroporated into R1 embryonic stem cells. Genomic DNAs from 288 G418-resistant clones were digested with BamHI and screened by Southern blot using 5′ probe. Eight homologous recombinants were obtained. Germline chimeras were generated from embryonic stem #144 by aggregation with d-2.5 embryonic 8 cells. The NEO cassette was deleted by crossing with Cre transgenic mice (Jackson ImmunoResearch Laboratories), and its deletion was confirmed by Southern blot and PCR (unpublished data). Offspring were backcrossed with BALB/c more than nine generations and typed as B7-DC WT and KO by Southern blot and PCR. B7-DC expression was completely abolished on BMDCs (Fig. 1), and spleen DCs (unpublished data). All experimental protocols were reviewed and approved by Johns Hopkins University Institutional Animal Care and Use Committee.

PE-conjugated anti-mouse B7-DC (TY25) and anti-mouse B7-H1 (M1H6) mAbs were purchased from eBioscience. PE-conjugated anti-mouse OVA-specific TCR (KJ1-26) mAb was purchased from Caltag Laboratories. Purified anti-mouse CD16/32 (2.4G2), IL-4 (BVD-1D11), and IFN-γ (R4-6A2) mAbs, PE-conjugated CD8 (Ly-2), CD80 (1G10), CD86 (GL1), Pan-NK cells (DX5), and I-A^d (AMS-32.1) mAbs, Cy-Chrome-conjugated CD45R/B220 (RA3-6B2) mAbs, and APC-conjugated CD3 (145-2C11), CD4 (RM4-5), and CD11c (HL3) mAbs were purchased from BD Biosciences. Recombinant mouse GM-CSF, IL-4, and IL-12 were purchased from R&D Systems.

For in vitro T cell assays, single-cell suspensions of spleen from B7-DC WT or KO mice were prepared by purification by Pan T cell isolation kit (Miltenyi Biotec) and cultured in complete medium with ConA for 72 h. The last 18 h before end of culture, ³H-thymidine was pulsed, and its incorporation was measured. For antigen-specific responses, CD4 T cells from DO11.10 mice were purified by using a CD4⁺ T cell isolation kit (Miltenyi Biotec) and labeled with CFSE before culture. The purity of CD4 T cells was monitored by FACS and was always >97%. BMDCs were prepared by culturing with recombinant mouse GM-CSF (1,000 unit/ml) in the presence or absence of IL-4 (1,000 unit/ml).

Cells were preincubated with unlabeled anti-CD16/32 mAb to avoid nonspecific binding of antibodies to Fc receptor and then incubated with PE- or APC-conjugated mAb. After washing twice with FACS buffer, the stained cells (live-gated on the basis of forward and side scatter profiles) were analyzed on a FACSCalibur (BD Biosciences).

1.0 × 10⁴ purified CD4 T cells from DO11.10 mice were cultured for 5 d with 3.0 × 10⁴ BMDCs and 0.3 μg/ml of OVA_323-339 peptide in the presence of recombinant IL-12 (1.0 unit/ml) and anti-mouse IL-4 mAb (10 μg/ml) for Th1 generation or with recombinant IL-4 (100 unit/ml) and anti-mouse IFN-γ mAb (10 μg/ml) for Th2 generation. 6 h before the end of culture, additional OVA_323-339 peptide (1.0 μg/ml) and GolgiStop (BD Biosciences) were added to each well for restimulation with or without purified anti-CD3 mAb (1.0 μg/ml). After surface CD4 was stained, cells were fixed and permeabilized by Cytofix/Cytoperm kit (BD Biosciences). Then intracellular IFN-γ and IL-4 were stained and analyzed by FACS.

5 million purified CD4 T cells from DO11.10 mice or CD8 T cells from clone 4 mice were labeled with CFSE and transferred to WT or B7-DC KO mice on d −1. For immunization, OVA_323-339 peptide– or HA_518-526 peptide– pulsed BMDCs were injected into WT or B7-DC KO mouse on d 0. Peptides were obtained from Colorado State University. Single-cell suspensions were prepared from each organ of immunized mice, cell numbers were counted, and CFSE dilution of CD4⁺KJ1-26⁺ T cells or CD8⁺Thy1.1⁺ T cells were analyzed by FACS. In vivo cytotoxic assay was performed by transferring CFSE^hi (1 μM)- and CFSE^lo (0.1 μM)-labeled Balb/c splenocytes. HA_518-526 peptide was pulsed only on the CFSE^lo population. The proportions of CFSE^hi and CFSE^lo cells were measured by FACS on d 5, and the cytotoxicity was calculated by the formula: % cytotoxicity = (% CFSE^hi − %CFSE^{hi control}) × 100 / (% CFSE^hi × % CFSE^{lo control}).

5 million purified DO11.10 CD4 T cells were transferred intravenously on d −1. Mice were immunized with 1 million OVA_323-339 peptide–pulsed BMDCs on d 0. 0.5 million CD4⁺KJ1-26⁺ T cells from spleen were sorted by FACS-Vantage (Becton Dickinson) at the Johns Hopkins Medical Institutions Oncology Center Cell Sorting Facility. The purity among live cells was >99%. Total mRNA from sorted cells was extracted, and first-strand cDNA was prepared by You-Prime First-Strand Beads (Amersham Biosciences). Real-time PCR reactions were performed in ABI PRISM 7700 Sequence Detection System (Applied Biosystems) as described (31, 32). Data were treated according to the manufacturer's instructions in User Bulletin #2. Briefly, standard curves were prepared by plotting the cycle threshold (Ct) by log ng input RNA. The inverse logs were calculated and experimental values were divided by 18s internal controls. Relative transcripts were determined by the formula: 1/2^{(CTtarget - CTcontrol)}.

Mice were injected with 100 μg NP¹⁷-KLH (Biosearch Technologies, Inc.) with alum (Pierce Chemical Co.) on d 0 and 14. Mice sera were collected on d 28. The titers of isotype and subclass of anti-NP–specific antibody were detected by ELISA using NP²⁵-BSA (Biosearch Technologies, Inc.) for antibody capture and SBA Clonotyping system/HRP kit (Southern Biotechnology Associates, Inc.) for detection (33).

For generating liver-specific tumor metastasis, 1 × 10⁵ cells of the Balb/c mouse–derived colon cancer cell line, CT26 were injected via the hemispleen technique on day 0 (34). Survival was monitored every day, and all dead mice were confirmed to possess liver metastases. To analyze antigen-specific CTLs after generating liver metastases of CT26, liver lymphocytes were extracted by the Percoll technique and stained with AH-1 peptide–pulsed MHC class I tetramer together with anti-mouse CD8 mAb.

Fig. S1 shows the flow cytometric analysis of lymphocyte subsets in thymus and spleen of B7-DC KO mice. Fig. S2 shows in vivo CD4 T cell proliferation in the liver after dendritic cell immunization.

This work was supported in part by a National Institutes of Health grant and gifts from William and Betty Topercer, Dorothy Needle, and the Commonwealth Foundation. This project was partly supported by Special Coordination Funds for Promoting Science and Technology from the Ministry of Education, Culture, Sports, Science and Technology, the Japanese Government. This project was partly supported by Uehara Memorial Foundation. Dr Pardoll is a Januey Scholar and holds the Seraph Chair for Cancer Research.

The authors have no conflicting financial interests.