Endothelial stroma programs hematopoietic stem cells to differentiate into regulatory dendritic cells through IL-10

Dendritic cells (DCs) play crucial roles in the initiation and regulation of immune responses.^1,2  The ability of DCs to initiate immune responses or induce tolerance is strictly dependent on their maturation state or subsets. It has been reported that immature DCs that are deficient of costimulatory molecules can induce T-cell anergy, generate regulatory T (Treg) cells, and promote alloantigen-specific tolerance. Several types of DCs with negative regulatory functions have been reported.³  Most regulatory DCs are prepared in vitro using immunosuppressive cytokines, such as IL-10 and TGF-β.^4-6  However, this may not reflect the real differentiation of regulatory DCs in the immune microenvironment in vivo.

The development of hematopoietic cells in vivo occurs in the context of microenvironment or niche,⁷  which consists of many types of stromal cells, such as fibroblasts, macrophages, endothelium cells, and adipose cells. The microenvironment provides various signals for hematopoietic cell development.⁷  Different microenvironments support different types of cell differentiation. The spleen is an important lymphoid organ, and the mouse spleen maintains hematopoietic function throughout life.⁸  Furthermore, hematopoietic stem cells (HSCs) are found in spleens of adult mice.⁸  Therefore, it is conceivable that HSCs in the spleen may differentiate into different immune cells in situ.

Splenic stromal cells cultured in vitro could mimic the splenic microenvironment in vivo to some extent, despite their differences in some constituents. There is evidence that long-term cultured splenic stromal cells can support the development of dendritic-like cells in the absence of exogenous cytokines, and the dendritic-like cells have the phenotype and function of DCs,^9-13  strongly suggesting that splenic microenvironment could induce hematopoietic progenitors to differentiate directly into DCs. Stromal cells cultured in vitro consist of multiple components. Purification of the various components will help to study the role of specific cell type in the induction of DCs. We established the method of preparing endothelial splenic stroma cells (ESSCs) and identified their characteristics, finding that ESSCs can drive mature DCs (mDCs) to proliferate and differentiate into regulatory DCs, whose natural counterpart exists in normal spleens.¹⁴  Therefore, we wondered whether its other origin in addition to mDCs; that is, could the splenic microenvironment differentiate hematopoietic progenitors to regulatory DCs? If so, what are its characteristics, and what is the mechanism? In this study, we show that ESSCs can differentiate HSCs into a distinct regulatory DC subset, CD11b^hiIa^lo DCs, which secret TGF-β, IL-10, and NO and have regulatory functions in vitro and in vivo. IL-10, secreted by ESSCs, plays a critical role in their differentiation. Its inhibitory function is mediated through NO but not through induction of Treg cells or T-cell anergy. These results suggest that splenic microenvironment may physiologically induce regulatory DC differentiation in situ.

Homozygous DO11.10 mice, heterozygous C57BL/6-TgN(ACTbEGFP)1Osb mice (GFP mouse), and IL-10 gene knock-out mice (H-2k^b) were purchased from the Jackson Laboratory (Bar Harbor, ME). Because of the low expression of Sca-1 on HSCs from Balb/c mice,¹⁵  (C57BL/6 × DO11.10) F1 mice were prepared by crossing C57BL/6 with DO11.10 mice. (GFP × DO11.10) F1 mice were obtained by crossing DO11.10 mice with C57BL/6-TgN (ACTbEGFP)1Osb mice.

RPMI-1640 medium (PAA Laboratories, Linz, Austria) was supplemented with 10% FCS (PAA Laboratories), 2 mM L-glutamine, 1% sodium pyruvate, and 2 × 10^–5 M β2 mercaptoethanol, all from Sigma-Aldrich (St Louis, MO). Recombinant mouse GM-CSF and IL-4 were purchased from PeproTech (London, United Kingdom). Fluorescein-conjugated mAbs to CD3e, CD4, CD8α, CD11b, CD11c, CD14, CD25, CD40, B220, Gr-1, CD69, CD80, CD86, CD117, Sca-1, I-A, KJ1-26, biotinylated Flt3 mAb, and isotype control mAbs were purchased from BD Pharmingen (San Diego, CA). Microbead-conjugated mAbs to CD4, CD8α, CD11b, CD11c, B220, CD117, FITC, and PE and the Lineage Cell Depletion Kit were from Miltenyi Biotec (Bergisch Gladbach, Germany). Fluorescein-conjugated mAbs to IL-2 and IFN-γ were obtained from eBioscience (San Diego, CA). Neutralization Abs to IL-10, TGF-β, M-CSF, and isotype control mAbs were purchased from R&D Systems (Minneapolis, MN). 2.4G2 for blocking mouse FcR I and II was purchased from BD Pharmingen. 7-AAD, lipopolysaccharide (LPS), saponin, brefeldin A, bovine serum albumin (BSA), 1-methyltryptophan, indomethacin, 1,4 PBIT, and CFSE were from Sigma (St Louis, MO). Rat sera were purchased from Gibco (Grand Island, NY) for blocking use.

ESSCs derived from newborn mice were prepared and maintained in long-term culture in RPMI1640/20% FCS by passaging to new plates each week.¹⁴ 

The strategy for HSC preparation was described by Christensen and Weissman.¹⁶  Briefly, cells from bone marrow were labeled with biotinylated mAbs against lineage antigens (CD5, B220, Gr-1, 7-4, and Ter-119) and Flt3 and were then incubated with streptavidin-PerCP–labeled antibody or antibiotin microbeads (for GFP⁺ mice to deplete Lin⁺Flt3⁺ cells by magnetic cell sorting [MACS]). These cells were further labeled with PE-conjugated anti-CD117 mAb and FITC or PE-Cy5 (for GFP⁺ mice) conjugated anti–Sca-1 mAb. Lin^–Flt3^–CD117⁺Sca-1⁺ cells were sorted as HSCs by fluorescence-activated cell sorting (FACS) Diva (Becton Dickinson, Mountain View, CA).

In clonal assay experiments, single Lin^–Flt3^–CD117⁺Sca-1⁺ HSCs from normal or GFP⁺ mice were directly seeded onto ESSC monolayers in 96-well flat plates by FACSDiva. Lin^–Flt3^–CD117⁺Sca-1⁺ cells (1 × 10³/mL) were added onto ESSC monolayers or polyformaldehyde-fixed ESSC monolayers in 24-well plates. After 10 days of culture, nonadherent cells were purified using anti-CD11b microbeads and collected as CD11b^hiIa^lo DCs. The purity was greater than 97%. Transwell experiments were performed in 24-well plates: 1 × 10³ Lin^–Flt3^–CD117⁺Sca-1⁺ HSCs in 300 μL culture medium were plated in the upper insert, separated from a coculture of ESSC monolayer by a 0.4-μm permeable membrane (a total culture volume of 1 mL). In some experiments, large numbers of CD11b^hiIa^lo regulatory DCs were needed, so Lin^–CD117⁺ bone marrow progenitors, prepared with the Lineage Cell Depletion Kit and anti-CD117 microbeads, were cocultured with ESSC monolayer.

Bone marrow–derived DCs (BMDCs) were generated from bone marrow cells in the presence of GM-CSF and IL-4 using established protocols.^17-20  For immature DCs (imDCs), on day 4, the dendritic proliferation clusters were collected and purified by anti-CD11c micobeads.²¹  Purified imDCs under the stimulation of LPS (0.01 μg/mL, for 4 days) were used as mDCs. The purity was greater than 95%.

CD4 T cells were obtained by negative selection using anti-CD11b and anti-CD11c microbeads and then positive selection using anti-CD4 microbeads. The purity was greater than 96%.

To isolate the in vivo natural counterpart of CD11b^hiIa^lo DCs, CD4^–CD8^–B220^– splenocytes were collected by negative selection with MACS. The negative cells were further labeled with PE-Cy5–conjugated anti-CD11c mAb, PE-conjugated anti-CD11b mAb, and FITC-conjugated anti-Ia mAb. Next, the cells with CD11c low expression were gated, and then CD11c^loCD11b^hiIa^lo cells were sorted by FACSDiva.

Cell surface marker and intracellular cytokine detection, proliferating T-cell counts, and analysis of phagocytosis ability were carried out as we described before.¹⁴  To test T-cell proliferation in 96-well plates after 5 days of culture, 120 μL/well culture supernatants were collected for cytokine assay by enzyme-linked immunosorbent assay (ELISA), then FITC-conjugated anti-CD4 mAb and 7AAD in 10 μL PBS were added. Alternatively, in some experiments, PE-conjugated anti–KJ1-26 mAb was also added and incubated for 15 minutes at 4°C. Before being collected for analysis by FACSCalibur (Becton Dickinson), 160 μL PBS per well was added to bring the final volume of each well up to 250 μL. Finally, cells were collected at high speed for 52 seconds, and the numbers of CD4⁺7AAD^– or CD4⁺KJ1-26⁺7AAD^– cells were counted.

Naive or activated (by mDCs for 24 hours in the presence of OVA_323-339 peptide) CD4 T cells (1 × 10⁵) were cultured in 96-well round-bottom plates with 1 × 10⁴ APCs (imDCs, mDCs, or CD11b^hiIa^lo DCs) in the presence of or absence of CD4 T cells (1 × 10⁵) primed by CD11b^hiIa^lo DCs; after 5 days of culture, viable cell numbers were counted by FACS. In some experiments, 1-methyltryptophan, indomethacin, 1,4 PBIT, neutralizing anti–IL-10 mAbs or anti–TGF-β1, -β2, and -β3 mAbs were added at the beginning of the culture.

Cytokines produced in culture supernatants at day 5 from mixed lymphocyte reaction (MLR) were detected by using ELISA kits for IL-2 and IFN-γ. For detection of IL-12p70, IL-10, and TGF-β produced by APCs, cells were cultured for 24 hours with or without LPS stimuli (0.5 μg/mL). NO production was measured as the nitrite concentration using the Griess assay.¹⁴ 

Naive DO11.10 T cells (6 × 10⁶) were transferred intraperitoneally into BALB/c mice. APCs (6 × 10⁶) were incubated with 1 μM OVA_323-339 peptide at 37°C for 6 hours, washed, and injected intraperitoneally into DO11.10 T-cell–transferred mice. After 5 days of immunization, the single cells from mesenteric lymph nodes and spleens were double stained with FITC-conjugated anti-CD4 mAb, PE-conjugated anti-OVA_322-339 peptide–specific T-cell receptor (TCR) mAb KJ1-26, and were counted by FACS. The cells were stained with both F27C-CD4 and PE-KJI-26.

The in vivo suppression of allo-DTH by CD11b^hiIa^lo DCs was performed as reported.²²  Briefly, B6 mice were immunized on the dorsal flank by subcutaneous inoculation of BALB/c spleen cells (1 × 0⁷/mouse) on day 0 and challenged on day 7 and day 14 at the hind footpad by injecting the same antigens (1 × 10⁷/mouse). APCs, derived from B6 mice, were injected intraperitoneally (3 × 10⁶/mouse) on days –6, –4, 0, 3, and 6. Footpad thickness was then measured on day 8 and day 15, respectively, with a calipers-type engineers' micrometer by a third experimenter who did not know the sample identity. The extent of swelling was calculated as the thickness of the right footpad (receiving BALB/c spleen cells) minus the baseline thickness of the left footpad (receiving B6 spleen cells).

All experiments were performed at least 3 times. All data analysis was performed using a 2-tailed Student t test. P less than .05 was considered statistically significant.

Our study protocol was approved by the review board of Second Military Medical University.

ESSCs, as described previously, express endothelial cell marker CD106 and have endothelial cell morphology.¹⁴  Lin^–Flt3^–CD117⁺Sca-1⁺ HSCs were sorted out from mouse bone marrow by FACSDiva (Figure 1A) and cultured with ESSC monolayer for at least 10 days before analysis. During the culture, the small round HSCs proliferated and gradually became larger nonadherent dendritic cell-like (DC-like) cells with many long dendrites (Figure 1B), displaying high phagocytic capacity (Figure 1C) like imDCs. These DC-like cells express high levels of the myeloid lineage marker CD11b and low levels of CD11c without expressing any other lineage markers, including CD3, CD4, CD8, B220, Gr-1, and CD14 (Figure 1D). Instead, these DC-like cells express all of the functional markers of DCs, such as Ia, CD86, CD80, and CD40 at low levels, similar to imDCs (Figure 1E). However, in contrast to imDCs, such expression of these markers at low levels was not increased when the cells were stimulated by LPS (Figure 1E), indicating a stable immature-like phenotype for these cells. On the basis of the phenotypic profiles, we termed these DC-like cells CD11b^hiIa^lo DCs. Analysis of cytokine profiles revealed that CD11b^hiIa^lo DCs spontaneously secreted TGF-β at a higher levels than imDCs and mDCs. IL-10 and NO were also secreted by CD11b^hiIa^lo populations and were significantly enhanced by LPS stimulation (Figure 1F), suggesting that CD11b^hiIa^lo DCs maybe be involved in immune regulation.

Although it is reported that splenic stromal cell–induced DCs are potent stimulators for allogeneic and syngeneic MLR and for antigen-specific T-cell proliferation,¹¹  CD11b^hiIa^lo DCs had a very limited (in allo-MLR) or no (in auto-MLR) capacity to stimulate T-cell proliferation even when CD11b^hiIa^lo DCs were stimulated by 10 ng/mL LPS for 4 days (Figure 2A-B). Interestingly, imDCs had a similar capacity to stimulate T-cell proliferation in allo-MLR and auto-MLR as mDCs, because live imDCs in this system may be induced to mature by T cells (Figure 2A-B ). These results demonstrate that CD11b^hiIa^lo DCs were a distinct subset of DCs with stable immature-like phenotype, different from imDCs.

We further characterized the capacity of CD11b^hiIa^lo DCs in antigen presentation to T cells. As shown in Figure 2C, compared with mDCs and imDCs, CD11b^hiIa^lo DCs could only make 30% to 40% of T cells express the early activated markers of T cells. Furthermore, we found that CD11b^hiIa^lo DCs could promote peptide-specific CD4 T cells to secrete high levels of IFN-γ, similar to mDCs (Figure 2D). We postulated that this distinct DC subset, CD11b^hiIa^lo DCs, which had a stable immature-like phenotype, could spontaneously secret TGF-β, and also had a powerful capacity to stimulate T cells to secrete high levels of IFN-γ, may have special functions in immune regulation, because it has been shown that IFN-γ plus NO can inhibit T-cell proliferation.²³ 

As shown in Figure 1, CD11b^hiIa^lo DCs had a stable immature-like phenotype and secrete TGF-β, an important cytokine with negative regulation of immune response.^23,24  We postulated that CD11b^hiIa^lo DCs may have regulatory functions. To determine this, we added CD11b^hiIa^lo DCs into CD4/mDCs coculture systems to see whether CD11b^hiIa^lo DCs could inhibit T-cell proliferation. As shown in Figure 3A, CD11b^hiIa^lo DCs, not imDCs, significantly suppressed the peptide-specific proliferation of naive peptide-specific CD4 T cells stimulated by mDCs. Similar results were observed in CFSE-labeling experiments (Figure 3B). However, the concentrations of IL-2 and IFN-γ in culture supernatant and levels of intracellular expression of IL-2 and IFN-γ were not reduced in the presence of CD11b^hiIa^lo DCs (Figure 3C-D).

Then we wondered whether CD11b^hiIa^lo DCs could suppress the proliferation of activated T cells that had been primed by mDCs. To determine this, naive peptide-specific CD4 T cells were stimulated by mDCs for 24 hours, and then CD11b^hiIa^lo DCs were added into the cultures. As shown in Figure 3E, the proliferation of the activated T cells was also completely suppressed by CD11b^hiIa^lo DCs, suggesting that the suppressive function of CD11b^hiIa^lo DCs may not be in the primary phase of T-cell activation.

To determine whether CD11b^hiIa^lo DCs had the ability to suppress the proliferation of T cells in vivo, purified naive peptide-specific CD4 T cells, together with various APCs that had been pulsed with OVA_323-339 peptide for 6 hours in vitro, were transferred into the peritoneal cavities of BALB/c mice. After 5 days, KJ1-26–positive cells from draining lymph nodes and spleens of mice that had received the transfers were detected with FACS. As shown in Figure 3F, the proliferation of DO11.10 T cells was significantly suppressed by CD11b^hiIa^lo DCs.

Taken together, the results showed that CD11b^hiIa^lo DCs have a powerful ability to suppress T-cell proliferation both in vitro and in vivo, demonstrating that CD11b^hiIa^lo DCs are a distinct subset of DCs with regulatory function.

Because CD11b^hiIa^lo DCs were potent in inhibiting the proliferation of CD4 T cells, which would suppress the immune response, we wondered whether CD11b^hiIa^lo DCs could suppress allospecific immune reactions in vivo. Allo-DTH experiments were performed for the allo-DTH experiment's simplicity and reliability to study allospecific immune reactions.²⁵  As shown in Figure 4, the footpad swelling of B6 mice receiving the alloantigen (BALB/c splenocytes) was suppressed more significantly by infusion with CD11b^hiIa^lo DCs derived from B6 mice. These results suggested that CD11b^hiIa^lo DCs could be used as a kind of negative regulator for immune response, thus outlining one approach to prevent or treat transplantation rejection or autoimmune diseases.

There is evidence showing that suppressive effects of DCs on T-cell proliferation were indirectly through the induction of Treg cells,^4-6  with an immunosuppressive function and cytokine profiles distinct from either Th1 and Th2 T cells,²⁶  or T-cell anergy, characterized by growth arrest state following an antigen encounter.²⁷  To test whether CD11b^hiIa^lo DCs induce Treg cells or cause T-cell anergy, we cocultured naive (DO-11.10 × GFP) F1GFP⁺ CD4 (GFP⁺CD4) T cells with CD11b^hiIa^lo DCs (from F1 mice) in the presence of OVA_322-339 peptide. On day 7, these CD11b^hiIa^lo DC-induced GFP⁺CD4 T cells purified from the coculture system by microbeads conjugated with anti-CD4 mAbs or naive (DO-11.10 × GFP) F1GFP^–CD4 (GFP^–CD4) T cells were stimulated by mDCs at the same time. As Figure 5A shows, CD11b^hiIa^lo DC-induced GFP⁺CD4 T cells displayed a proliferation ability similar to that of naive GFP^–CD4 T cells, showing that CD11b^hiIa^lo DC-induced GFP⁺CD4 T cells were not anergic. Similarly, if the purified CD11b^hiIa^lo DC-induced GFP⁺CD4 T cells were added into the mDC/GFP^–CD4 coculture system, the proliferation of naive GFP^–CD4 T cells was not suppressed in the presence of CD11b^hiIa^lo DC-induced GFP⁺CD4 T cells (Figure 5B), demonstrating that CD11b^hiIa^lo DCs did not induce Treg cells, which had the capacity to suppress other T-cell proliferation. These results suggested that there may be other mechanisms involved in the inhibitory function of CD11b^hiIa^lo DCs.

Because CD11b^hiIa^lo DCs did not induce the production of Treg cells or make T-cell anergy, we postulated that maybe other factors were involved in this process. First, to determine whether this inhibitory effect was mediated by direct contact or by soluble factors, the effects of fixed CD11b^hiIa^lo DCs or CD11b^hiIa^lo DC supernatant were detected. As shown in Figure 5C, the proliferation of CD4 T cells were significantly suppressed by CD11b^hiIa^lo DC supernatant, not by fixed CD11b^hiIa^lo DCs, suggesting that the inhibitory effect of CD11b^hiIa^lo DCs was dependent on the autocrine soluble factors.

Next, we tried to find out what factors in the supernatant were involved in this process. IL-10 and TGF-β produced by CD11b^hiIa^lo DCs are 2 cytokines that are well known for inhibiting the proliferation of CD4 T cells.^28,29  However, addition of blocking anti–IL-10 and anti–TGF-β antibody did not reverse the inhibitory effects of CD11b^hiIa^lo DCs (Figure 5D).

It has been reported that PGE₂³⁰  and indoleamine 2,3-dioxygenase (IDO; a tryptophan-catabolizing enzyme)³¹  can inhibit T-cell proliferation, so we added indomethacin, an inhibitor of PGE₂ biosynthesis or 1-methyltryptophan, a specific IDO inhibitor, into a CD4/mDC/CD11b^hiIa^lo DC coculture system. We found that the inhibitory effects of CD11b^hiIa^lo DCs were not substantially reversed (Figure 5E).

It has become increasingly evident that NO is involved in inhibition of T-cell proliferation.³²  The high levels of NO secreted by CD11b^hiIa^lo DCs (Figure 1F) suggest that NO may be involved. We added series of concentrations of NOC-18, an NO donor, into a CD4/mDC coculture system at the beginning of culture, and found that the proliferation of CD4 T cells was inhibited in a dose-dependent manner. However, similar inhibitory effects were achieved by a much higher concentration of NO than that of NO in CD4/mDC/CD11b^hiIa^lo DC coculture systems (Figure 5F). Furthermore, as shown in Figure 5G, the CD4 T-cell proliferation suppressed by CD11b^hiIa^lo DCs could be significantly restored by 5 μg/mL 1,4 PBIT, an inhibitor of iNOS, which reduces the concentration of NO to levels found in CD4/mDC groups. Increasing the concentration of 1,4 PBIT to 10 μg/mL did decrease the NO significantly but failed to further restore the proliferation of CD4 T cells, and increasing the concentration to 20 μg/mL resulted in toxicity to CD4 T cells. These results demonstrated that NO was indeed involved in suppressive effects of CD11b^hiIa^lo DCs but not in a dependent manner, because completely inhibiting NO production only partly restored the suppressive effects, suggesting there may be novel factors that we do not yet know in the supernatants.

What factors determine the differentiation of Lin^–Flt3^– CD117⁺Sca-1⁺ bone marrow HSCs into regulatory CD11b^hiIa^lo DCs? First, to test whether this differentiation process depends on direct contact with ESSCs or soluble factors, Lin^–Flt3^– CD117⁺Sca-1⁺ HSCs were cultured with fixed ESSCs or in the upper insert of a transwell system, separated from ESSCs by permeable membrane. As shown in Figure 6A, only CD11b^hiIa^lo DCs derived from Lin^–Flt3^–CD117⁺Sca-1⁺ HSCs cultured in the transwell system (CD11b^hiIa^lo-trans) had the capacity to inhibit the proliferation of CD4 T cells. These results demonstrated that CD11b^hiIa^lo DC differentiation was dependent on soluble factors from ESSCs.

To further understand which factor was involved in CD11b^hiIa^lo DC differentiation, TGF-β, IL-10, and M-CSF, secreted by ESSCs,¹⁴  were blocked by their respective neutralizing antibody. After 10 days of culture, different differentiated DCs were detected by their inhibitory functions. As shown in Figure 6B, blocking IL-10, instead of TGF-β or M-CSF, in the coculture system led to partial loss of inhibitory functions of CD11b^hiIa^lo DCs, indicating IL-10 was involved in the process. To confirm this, ESSCs from IL-10 gene knockout mice were used. Although the morphology (data not shown) and phenotype (Figure S1, available on the Blood website; see the Supplemental Figure link at the top of the online article) of the cells induced by IL-10–KO-ESSCs were similar to those of the cells induced by wild-type ESSCs, their inhibitory ability was significantly reduced. However, when exogenous IL-10 was added into the IL-10–KO-ESSC coculture system, the lost inhibitory ability was recalled (Figure 6C). These results demonstrate that IL-10 is indeed important in the differentiation of CD11b^hiIa^lo DCs from HSCs. This conclusion was further confirmed by another experiment, which showed that the yield of CD11b^hiIa^lo DCs induced by IL-10–KO-ESSCs was significantly lower than that of CD11b^hiIa^lo DCs induced by normal ESSCs (Figure 6D).

To determine whether the natural counterparts of CD11b^hiIa^lo DCs exist in vivo, we first analyzed the splenocytes with anti-CD11b, anti-Ia, and anti-CD11c mAbs on the basis of the phenotype of CD11b^hiIa^lo DCs and found that the CD4^–CD8^–B220^–CD11c^loCD11b^hiIa^lo populations do exist in the spleen, and were about 11.7% of all CD4^–CD8^–B220^–CD11c⁺ DCs in spleens of normal mice. Then these cells were sorted to test their functions (Figure 7A). The peptide-specific CD4 T-cell proliferation was significantly suppressed by CD11b^hiIa^lo DCs that were directly sorted from the mouse spleen, although the suppressive capacity of this natural counterpart could not be comparable with that of CD11b^hiIa^lo DCs induced in vitro. This might contribute to the difference between cultured cells in vitro and freshly isolated cells in vivo. Figure 7B also showed that neither the freshly sorted CD11b^hiIa^lo DCs (in vivo) nor CD11b^hiIa^lo DCs (in vitro prepared) could stimulate CD4 T-cell proliferation.

We also found that the in vivo quantity of CD11b^hiIa^lo DCs in IL-10 KO mice was significantly lower than that of CD11b^hiIa^lo DCs in normal mice (Figure 7A), convincingly demonstrating that IL-10 contributes to the development of CD11b^hiIa^lo DCs in vivo. Because CD11b^hiIa^lo DCs (in vivo) in IL-10–KO mice were too few to be isolated, we cannot compare their function with CD11b^hiIa^lo DCs sorted from normal mice.

In this work, we demonstrate that splenic stroma induces bone marrow HSCs to differentiate into regulatory DCs that are characterized by high expression of CD11b and low expression of CD11c and MHC II molecules, so these DCs were designated CD11b^hiIa^lo DCs. CD80, CD86, and CD40 are stably expressed on its surface at low levels. Because LPS stimulation could not increase their expression, CD11b^hiIa^lo DCs are regarded as some kind of mDCs, instead of the common DCs in an immature state. CD11b^hiIa^lo DCs display powerful phagocytic capacity, appearing to contradict the conventional view on DCs' maturation, but there are several works indicating that even mDCs retain some capacity to endocytosis and present antigens.^33-35  Moreover, we found that CD11b^hiIa^lo DCs secrete high levels of TGF-β and IL-10, characteristic of tolerogenic/regulatory DCs.^4-6  Indeed, our results show that CD11b^hiIa^lo DCs have the capacity to inhibit the peptide-specific proliferation of CD4 T cells stimulated by mDCs both in vitro and in vivo, demonstrating that, functionally, CD11b^hiIa^lo DCs are a subset of regulatory DCs.

There are several kinds of regulatory DCs reported, most of them induced by cytokine cocktail in vitro. Although these are very simple and easy methods to study the origin of regulatory DCs, they cannot reflect the actual differentiation of regulatory DCs in vivo. The regulatory DCs reported until now have similar functions but different phenotypes. The CD11c and Ia low expression of CD11b^hiIa^lo DCs was different from that of CD11c^hiIa^hi regulatory DCs, induced by combination of mGM-CSF, mIL-10, and hTGF-β.⁴  Because CD80 and CD86 are expressed at low levels by CD11b^hiIa^lo DCs, it is clearly distinguished from the CD80^hiCD86^hiCD40⁺Ia^int DCs reported by Akbari et al.³⁶  Although CD11b^hiIa^lo DCs express CD45RB (data not shown), they do not express CD4, as expressed by CD11c^loCD45RB^hi DCs.⁶  Because CD11b^hiIa^lo DCs express low levels of CD80 and CD40 and are differentiated from HSCs, they are also different from the diffDCs that we have described that express high levels of CD80 and CD40 and are derived from mDCs.¹⁴  These results show that, phenotypically, CD11b^hiIa^lo DCs are a novel subset of regulatory DCs.

Functionally, inhibiting T-cell proliferation is the hallmark of regulatory DCs, but the molecular mechanisms of regulatory DCs are still poorly defined so far. There are at least 3 mechanisms that may account for their tolerogenic/regulatory properties, including absent costimulating signal, deletion of antigen-reactive cells by Fas/FasL system or tryptophan metabolites, and induction of Treg cells.³⁷  However, we demonstrate that the inhibitory function of CD11b^hiIa^lo DCs is not mediated by inducing the production of Treg cells or T-cell anergy, which was reported as the main inhibitory mechanism of regulatory DCs.^4-6  Although high levels of TGF-β and IL-10, which were shown to be important in the induction of tolerance^38-41  and Treg cells,^42,43  were secreted by CD11b^hiIa^lo DCs in our experiments, they did not promote the production of Treg cells. We thought this might contribute to the distinct functional features of CD11b^hiIa^lo DCs, which are induced by splenic stroma, a very complex and physiologic condition, not simply under the conditions of single or cocktail cytokines. Also, FasL was not detected on the surface of CD11b^hiIa^lo DCs (data not shown). Interestingly, we found that CD11b^hiIa^lo DCs could effectively present antigen to peptide-specific CD4 T cells and vigorously activate them to secrete high levels of IFN-γ without proliferation, potently inhibiting the proliferation of peptide-specific CD4 T cells stimulated by mDCs, without depressing IL-2 or IFN-γ secretion. These results showed that regulatory functions of CD11b^hiIa^lo DCs for T-cell response are multiple. Our further results suggest that NO was involved in this process, instead of IL-10, TGF-β, PGE2, or IDO.³¹  NO inhibits T-cell proliferation by acting on multiple members of the IL-2 signaling pathway, particularly JAK3 and JAK1.⁴⁴  These results clearly showed that CD11b^hiIa^lo DCs display their inhibitory function in a novel way: by directly secreting soluble inhibitory factors, including NO, instead of indirectly, such as by inducing Treg cells, confirming that, functionally, CD11b^hiIa^lo DCs are a distinct regulatory DCs subset.

Interestingly, a regulatory DC, CD11c^loCD45RB⁺, also induced by splenic stroma, was recently reported by Svensson et al⁴⁵ ; there are several differences between those cells and the CD11b^hiIa^lo DCs described here. First, CD11c^loCD45RB⁺ DCs were derived from lineage-negative c-kit⁺ progenitor cells, whereas CD11b^hiIa^lo DCs were from HSCs. Second, CD11b^hiIa^lo DCs were induced by endothelial stroma instead of by stroma comprising principally fibroblasts and macrophages, as reported by Svensson et al.⁴⁵  The most important difference is the inhibitory effect of CD11b^hiIa^lo DCs was NO involved, instead of inducing Treg cells. All the differences showed that CD11c^loCD45RB⁺ DCs and CD11^hiIa DCs are not the same subsets of regulatory DCs.

In one study, Kamath et al³⁴  found a rapid turnover of DC subsets in mouse spleen within 3 to 4 days, suggesting that in adult mice, the differentiation of DCs from progenitors in situ in the spleen might exist. But it is unclear how splenic microenvironment induces DC differentiation. Our results demonstrate that, in vitro, the ESSC-derived soluble factors are essential for the differentiation of CD11b^hiIa^lo DCs, and we further demonstrate that IL-10 secreted by ESSCs is important in this process by using IL-10–blocking antibody and IL-10–KO-mice. Although it is very difficult to directly demonstrate HSC differentiation into DCs in the spleen in situ from the beginning of development in vivo, we cannot exclude this possibility, because the spleen has been the hematopoietic organ in new neonatal phase and HSCs can be found even in adult spleens.⁸  Maybe our system mimicked this development process.

Moreover, natural counterparts of CD11b^hiIa^lo DCs that do not express CD4 and CD8 were found in vivo. Although CD4^–CD8^– DCs in spleens were reported,⁴⁶  it is unclear whether subpopulations in these double-negative populations exist, and how they function is unclear. In this report, we clearly showed that CD11c-low expression population in CD4^–CD8^– DCs can be further classified into several subsets according to the extent of CD11b and Ia expression. Interestingly, CD11b^hiIa^lo DCs had inhibitory function similar to that of CD11b^hiIa^lo DCs induced by splenic stroma in vitro, indicating that these CD11c^loCD11b^hiIa^lo DCs in normal mouse spleens were the natural counterparts of CD11b^hiIa^lo DCs in vitro and were a functionally specialized DC subset in the spleen. Our work, together with the findings of Wakkach et al,⁶  demonstrated that there are specialized regulatory DCs in the spleen that might play important roles in maintenance of immune balance.

We also demonstrate that in vitro–cultured CD11b^hiIa^lo DCs have the ability to suppress the proliferation of the peptide-specific CD4 T cells in vivo and to suppress alloantigen-specific responses in vivo, opening new preventive or therapeutic perspectives for the use of CD11b^hiIa^lo DCs in autoimmune disease, inflammatory disease, or allogeneic transplantation.

In conclusion, we demonstrate that the spleen microenvironment could self-regulate the immune response by driving HSCs in spleen to differentiate into CD11b^hiIa^lo DCs, a novel regulatory DC subset in an IL-10–dependent way. Such DCs secret TGF-β, IL-10, and NO and exhibit regulatory functions both in vitro and in vivo. The inhibitory function is NO involved, instead of inducing Treg cells or anergy. The ability of CD11b^hiIa^lo DCs to suppress allo-DTH in vivo indicates that these cells may have preventive or therapeutic prospects for some immunologic disorders.

We thank Ms Qian Li, Dr Sheng Xia, Ms Rui Zhang, and Ms Chunfang Luo for their excellent technical assistance. We thank Prof You-Wen He from Duke University for helpful discussion.