CD8α⁺ and CD8α⁻ Subclasses of Dendritic Cells Direct the Development of Distinct T Helper Cells In Vivo

Balb/c mice were purchased from Iffa-Credo. Balb/c IL-12 p40^−/− mice were provided by Dr. J. Magram (Hoffmann-La Roche, Nutley, NJ [5]). All mice were maintained in our pathogen-free facility and used at 7–9 wk of age.

DCs were purified as shown previously (6), except that spleen cells were digested with collagenase, further dissociated in Ca²⁺-free media in the presence of EDTA, separated into low- and high-density fractions on a Nycodenz gradient, and cultured overnight with 15 ng/ml rmGM-CSF, as described (3). DCs were pulsed with antigen (30 μg/ml KLH) during overnight culture (6). The CD8⁻ or CD8⁺ status was maintained after culture in GM-CSF–containing medium for 1–5 d (7; and our unpublished observations), suggesting that CD8α is a stable lineage marker. After overnight culture, nonadherent cells (containing at least 90% DCs, as assessed by morphology and specific staining, using anti-CD11c mAb, N418; reference 8) were incubated with anti-CD8α–coupled microbeads and separated according to CD8α expression by two passages over a MACS^® column (Miltenyi Biotec). The CD8α⁻ DCs were further enriched by incubation with anti-CD11c–coupled microbeads and positive selection over a MACS^® column. Alternatively, nonadherent cells collected after overnight culture were separated according to CD8 expression by FACS^® sorting: in brief, cells were double stained for CD11c expression using FITC-conjugated N418 and for CD8α expression using biotin-conjugated anti-CD8α mAb (PharMingen) followed by PE-streptavidin. The cells were gated based on characteristic forward and side light scatter, and two populations (CD11c⁺CD8α⁺ and CD11c⁺CD8α⁻) were sorted on a FACSVantage^® (Becton Dickinson). The proportion of CD8α⁺ to CD8α⁻ DCs at the end of the purification steps was 10–15% in all experiments performed.

Low-density spleen cells (see above) were enriched for CD11c expression and further separated according to CD8α expression using a Multisort anti-FITC kit (Miltenyi Biotech). Cells were cultured overnight with pansorbin (20 μg/ml; Calbiochem) plus IFN-γ and GM-CSF (20 ng/ml each), and the supernatant was assayed for IL-12 p70 using ELISA from Genzyme. The detection limit was 8 pg/ml.

Antigen-pulsed DCs were washed in RPMI 1640 and administered at a dose of 3 × 10⁵ cells into the hind footpads, according to a protocol described by Inaba et al. (9). When indicated, some groups of animals were treated with 1 mg anti-CD4 mAbs (GK 1.5), to selectively deplete CD4⁺ T cell subset in vivo, as described previously (10). Some mice were injected daily with 0.2 μg i.p. rmIL-12 on days 0, 1, 2, and 3. Draining popliteal LNs were harvested 5 d after DC injection.

LN cells were cultured in Click's medium supplemented with 0.5% heat-inactivated mouse serum and additives. The proliferation was measured as thymidine incorporation during the last 16 h of a 4-d culture. Culture supernatants were assayed for IL-2 after 24–48 h, and for IFN-γ, IL-4, IL-5, and IL-10 after 96 h of incubation. IL-2, IFN-γ, and IL-10 were measured as described (11). IL-4 and IL-5 were quantitated by two-site ELISA from Genzyme and PharMingen, respectively. The detection limits were 0.05 U/ml for IL-2, 1 ng/ml for IFN-γ, 15 pg/ml for IL-4, 2 U/ml for IL-5, and 0.3 ng/ml for IL-10.

DCs were purified from spleens, pulsed with KLH during overnight culture (6), and further separated according to CD8α expression by FACS^® sorting or by positive/negative selection on MACS^®. Reanalysis of the sorted cell populations confirmed purity >99% (FACS^®) or 97% (MACS^®). 3 × 10⁵ DCs were injected into the hind footpads of syngeneic mice, and the popliteal LNs were harvested 5 d later. The data in Fig. 1 a indicate that administration of purified CD8α⁺ or CD8α⁻ DCs, or both, loaded ex vivo with antigen, resulted in T cell priming, as assessed by KLH-dependent proliferation in culture. The proliferative response of LN cells from mice injected with CD8α⁺ DCs was consistently higher compared with animals primed with CD8α⁻ DCs. T cell priming was prevented by treatment of mice with neutralizing anti-CD4 mAbs, showing that the KLH-specific response was dependent on CD4⁺ T lymphocytes.

We next analyzed the cytokines (12) released by LN cells from mice primed with either subset in vivo. The data in Fig. 1 indicate that the subclasses of DCs have the potential to differentially skew cytokine production towards Th1 and Th2 tendencies (12): CD8α⁻ DCs induced the activation of cells secreting high levels of IL-4, IL-5, and IL-10 and low levels of IL-2 and IFN-γ, whereas CD8α⁺ DCs sensitized cells producing IL-2 and IFN-γ, but little IL-4, IL-5, and IL-10. Unseparated splenic DCs (11; and data not shown) or a combination of both subsets (at a proportion of 1 CD8α⁺ to 10 CD8α⁻) induced the activation of helper cells secreting a large array of lymphokines.

These results indicate that both CD8α⁺ and CD8α⁻ subsets of DCs can act as adjuvant of the immune response and differentially regulate the development of CD4⁺ Th cells. As in vitro data suggested that CD8α⁺ DCs could kill Fas-expressing cells (3), T lymphocytes may undergo Fas-mediated apoptosis once they are activated, i.e., later during the primary response. Therefore, we tested whether an anamnestic response was developed. Unseparated, KLH-pulsed DCs were injected into the hind footpads of mice that were untreated or immunized 14 d earlier with various DC populations. LN cells were harvested 2 d later and cultured with or without antigen. The data in Fig. 2 a indicate that a memory response was induced in all preimmunized groups, as assessed by antigen-specific T cell proliferation of LN cells from mice that received two injections of DCs, but not in groups that received only one injection. Of note, the cytokine profiles were determined by the subclass of DCs used to prime animals: mice injected with CD8α⁺ or CD8α⁻ DCs and boosted with unseparated DCs displayed a secondary response of Th1 and Th2 type, respectively (Fig. 2). These findings indicate that both DC subsets induce the development of memory helper cells that upon recall with the same antigen differentiate into distinct helper populations.

There is evidence that the maturation of Th precursors into biased Th1 or Th2 populations is strongly influenced by cytokines in the environment (13). In particular, IL-12 appears as the dominant cytokine driving the differentiation of Th1 lymphocytes in vitro and in vivo (14, 15). We found that CD8α⁺ DCs produced high levels of IL-12 heterodimer upon stimulation, whereas CD8α⁻ DCs secreted little if any IL-12 (Table I). The role of IL-12 in Th1 priming was further documented by the observation that CD8α⁺ DCs isolated from mice deficient for IL-12 (5) induced little IFN-γ and intermediate levels of IL-4 when injected into syngeneic mice, compared with CD8α⁺ DCs from wild-type mice (Fig. 3, a and b). Conversely, coinjection of rIL-12 and antigen-pulsed CD8α⁻ DCs resulted in the development of a polarized Th1-type immune response (Fig. 3, c and d).

There is increasing evidence that protection against parasites or infectious diseases relies on the character of the immune response, i.e., the Th1/Th2 balance. In particular, Th1 cells are important effectors involved in the eradication of intracellular infectious pathogens, whereas Th2 lymphocytes are efficient to eliminate extracellular parasites. Importantly, the development of the inappropriate Th subset not only fails to eradicate the pathogen but can cause immunopathology (for a review, see reference 13). Therefore, it is of major interest to identify the factors that influence the differentiation of distinct Th cell subsets in vivo. We show herein that two subclasses of DCs differentially regulate the development of Th cells secreting discrete sets of lymphokines: CD8α⁺ DCs direct the differentiation of Th1 cells, whereas CD8α⁻ DCs induce Th2-type responses in vivo. We further show that both DC subclasses induce memory responses, and that the activation of Th1 responses by CD8α⁺ DCs correlates with their production of IL-12 p70 heterodimer.

Our data are in agreement with recent reports that CD8α⁺ DCs are the source of IL-12. In Flt3 ligand–treated mice, a DC subset containing a majority of CD8α⁺ cells was shown to secrete very high levels of IL-12 (16). Similarly, most DCs exposed to soluble Toxoplasma gondii tachyzoite extract in vivo that stained for IL-12 p40 were shown to belong to the CD8α subset (17). In addition, our results suggest that the same cells, i.e., the transferred DCs, function as APCs and IL-12–producing cells, as DCs from p40 knockout mice have lost the capacity to induce a Th1-type response in a wild-type host. These observations and other studies suggest that IL-12 is a dominant factor in directing the development of Th1 cells producing high levels of IFN-γ in vitro and in vivo, although other factors may affect Th subset development (13). Of note, several reports suggest that IL-12 production by DCs requires their prior activation and that the amount of IL-12 depends on the mode of activation. Reis e Sousa and collaborators reported a CD40L- and IFN-γ–independent production of IL-12 after infection with T. gondii or injection of LPS (17). Koch et al. (18) found that ligation of CD40 and MHC class II molecules independently triggered IL-12 production by murine DCs. Similarly, CD40L was found to be the most effective stimulus to induce IL-12 by DCs generated from human peripheral blood monocytes (19). Therefore, it is tempting to speculate that, in our model, the transferred antigen-pulsed DCs are triggered to release IL-12 upon antigen-specific interaction with T lymphocytes in vivo. Experiments are underway to define the role of CD40– CD154 interaction in Th1 priming and to compare the IL-12 produced by DCs at various stages of maturation.

There is some evidence that CD8α⁺ and CD8α⁻ DCs belong to distinct lineages. Shortman and colleagues have found in intravenous transfer studies that CD8α serves as a marker of the DC progeny of the low CD4 precursor, in both the thymus and the spleen of irradiated recipients, thereby suggesting that CD8α⁺ DCs are of lymphoid origin (20, 21). Conversely, the CD8α⁻ DCs would be of myeloid origin, as they are relatives of monocytes and macrophages and are GM-CSF dependent (21, 7). Both classes of DCs maintain their CD8α⁻ or CD8α⁺ status in culture in the presence of GM-CSF, and therefore appear as stable distinct lineages (7). Of note, CD8α remained a stable marker on DCs cultured for 5 d in the presence of GM-CSF, IFN-γ, and/or activated T cells (our unpublished observations).

A recent study by Pulendran et al. (21a) confirmed that both subsets of DC differentially regulated the development of T helper cells in vivo. They showed that the lymphoid-related subsets induced high levels of IFN-γ and IL-2, but little Th2 cytokines, whereas the myeloid-related subset induced large amounts of IL-4 and IL-10, in addition to IFN-γ and IL-2. The CD8α⁺ population used in the present study is CD11c⁺CD11b^{dull or −}, and therefore is likely to represent the lymphoid-related population D/E described by Pulendran and colleagues (16). By contrast, the CD8α⁻ DCs are CD11c⁺CD11b⁺ and resemble the myeloid-related DC subset referred to as population C (data not shown). The distinct regulation of the IFN-γ and IL-2 synthesis, compared with our study, could be related to differences in the purification procedures, in the maturation state of the DCs transferred (fresh versus cultured DCs), in the form of antigen (peptide versus protein), and/or in the precursor frequency of antigen-reactive T cells (TCR transgenic versus wild-type mice). It is noteworthy that injection of CD8α⁺ and CD8α⁻ DCs isolated from mice treated with Flt3 ligand (provided by Dr. C. Maliszewski, Immunex, Seattle, WA) induced the development of similar Th cells compared with DCs purified from untreated mice, suggesting that administration of Flt3 ligand did not alter DC function (data not shown).

Two reports have shown that the cell population containing the highest proportion of CD8α⁺ cells was consistently less efficient at stimulating the proliferation of antigen-specific cells in vitro (3, 22). A ligand for Fas was demonstrated on the surface of CD8α⁺ but not CD8α⁻ DCs, and the suboptimal activation of T cells by CD8α⁺ DCs was associated with marked T cell apoptosis via Fas engagement (3). We show here that injection of pulsed CD8α⁺ and CD8α⁻ DCs induced equally strong T cell proliferative responses upon in vitro restimulation. Although we did not measure the expansion of T cells in situ, a difference between the in vitro and in vivo function of CD8α⁺ DCs could be due to the segregation of cell populations into distinct geographic compartments in vivo after T cell activation (23) compared with the confined microenvironment in culture plates. Alternatively, it is possible that Th1-type responses, which could be deleterious, are controlled by a feedback mechanism involving Fas- mediated killing of T lymphocytes once activated (24, 25). Interestingly, there is evidence that CD95L may be a mediator of costimulation and inflammation as well as a death agonist (26–30): CD95L has been shown to recruit neutrophils and activate their cytotoxic machinery, leading to local inflammation (29). As inflammatory products seem to induce the maturation of DCs and their migration to lymphoid tissues (31, 32), inflammation may be crucial for the initiation of immunity. Experiments are underway to test whether FasL expression by CD8α⁺ is required for the induction of a Th1-type response in vivo.

It is intriguing that CD8α⁺ and CD8α⁻ DCs seem to be located in distinct microenvironments of the spleen (16). In Balb/c mice, a majority of CD8α⁻ DCs reside at the margin between the red and white pulp, whereas most CD8α⁺ DCs are present in the zones where T cells are located (our unpublished observations). Injection of LPS results in the redistribution (32) of both subsets into the T cell area (our unpublished observations), suggesting that both subsets have migratory properties. It is notable that CD8α⁺ DCs express high levels of DEC-205, a multilectin-like receptor (33) which could be specific for carbohydrates that are common constituents of microbial cell walls. In addition, the expression of CD1d, an MHC-like molecule which presents antigens mainly derived from prokaryotes (34), has been shown to be highest on the CD8α⁺ DC subset (16). Therefore, it is tempting to speculate that the CD8α⁺ subset of DCs preferentially capture and present microbial antigens and elicit a Th1-type response.

In conclusion, two subsets of DCs, which differ by phenotype, functional, and histologic parameters, exist in the mouse, each one initiating a different class of response in vivo and thereby stimulating different effector mechanisms. Our data show that CD8α⁺ DCs drive the development of Th1-type immune responses, whereas CD8α⁻ DCs induce the differentiation of Th2-type responses. These observations predict that CD8α⁺ DCs might not be exploited to induce tolerance in vivo and confer immune privilege to grafts, but instead may be attractive for eliciting therapeutic antitumor immunity.

We thank Dr. K. Shortman for interesting discussions; Drs. B. Pulendran and C. Maliszewski for sharing unpublished results and for useful comments; Dr. M. Goldman for careful review of the manuscript; Dr. J. Magram for providing IL-12–deficient mice; and G. Dewasme, M. Swaenepoel, F. Tielemans, and P. Veirman for technical assistance.