CX₃CR1⁺c-kit⁺ Bone Marrow Cells Give Rise to CD103⁺ and CD103⁻ Dendritic Cells with Distinct Functional Properties¹

APCs are an essential component of the immune system that connects the innate and the adaptive immune responses. They are involved in maintenance of tissue homeostasis, inflammation, and the organization of the immune response (1).

Dendritic cells (DC)³ reside in peripheral tissues, such as the lung, and migrate under inflammatory as well as steady-state conditions to the regional lymph node, reflecting the continuous turnover process of these cells (2). Blood monocytes have been identified as a cell population that continuously replenishes those DC that left peripheral organs during turnover. CCR2^highLy6C^highCX₃CR1^low and CCR2^lowLy6C^lowCX₃CR1^high cells are the two main populations of blood monocytes in mice (3). Recent findings demonstrated that CCR2^highLy6C^highCX₃CR1^low and CCR2^lowLy6C^lowCX₃CR1^high cells are able to preferentially give rise to pulmonary CD103⁺ and CD103⁻ DC, respectively (4, 5).

CD103⁺ DC are also found in the intestinal lamina propria, mesenteric lymph nodes (LN), and skin-draining LN and have been described as the major DC subset of the lung (6, 7, 8). We have previously reported that CD103⁺ and CD103⁻ DC residing in the lung-draining bronchial LN (brLN) have evolved to acquire opposing functions in presenting innocuous inhaled Ag. Thus, under tolerogenic conditions, the CD103⁻ DC present innocuous Ag to CD4⁺ T cells, while the CD103⁺ DC, which do not express CD8α, are specialized in presenting Ag exclusively to CD8⁺ T cells (9). Considering that in the lung, CD8α DC are virtually missing and that most DC colonizing this tissue express CD11b, we hypothesized that under homeostatic as well as inflammatory conditions bone marrow progenitors might contribute to the replenishment of CD103⁺ and CD103⁻ DC in the lung.

To test this idea, we took advantage of a mutant mouse, in which enhanced GFP (eGFP) gene was knocked into the fractalkine receptor gene locus (CX₃CR1). CX₃CR1 is primarily expressed by subpopulations of NK cells, monocytes, macrophages, and DC but is also found on a subpopulation of bone marrow progenitor cells (10). Using bone marrow from CX₃CR1^+/gfp mice, we sorted lineage (lin)⁻CX₃CR1⁺c-kit⁺CD11b⁺/CD11b⁻ (GFP⁺c-kit⁺) as well as lin⁻CX₃CR1⁺c-kit⁻ (GFP⁺c-kit⁻) cells that were used as a source of myeloid progenitors for the in vitro and in vivo differentiation of DC. We demonstrate that CD103⁻ as well as CD103⁺ DC can be generated from both GFP⁺c-kit⁺ and GFP⁺c-kit⁻ cells cultured in vitro in the presence of GM-CSF. CD103⁻ DC displayed up-regulated transcripts for genes involved in innate immunity, whereas those involved in costimulation were down-regulated when compared with CD103⁺ DC, suggesting distinct functional activities of these two DC subsets. In vivo adoptive transfer of GFP⁺c-kit⁺ cells to irradiated recipients undergoing an allergic immune response allowed the induction of donor-derived CD103⁺ DC in the lung and the lung-draining LN. These results indicate that these progenitors can be recruited to sites of inflammation where they further differentiate in situ to DC, which then migrate to draining LN to induce adaptive immunity.

Mice expressing eGFP instead of the CX₃CR1 gene (CX₃CR1^+/gfp) on the CD45.2 C57BL/6 (B6) background have been described elsewhere (10). CD45.1 recipients, OT-I Thy1.1 and OT-II Ly5.1, mice that carry a transgenic TCR for the H2-K^b-restricted SIINFEKL peptide derived from OVA_257–264 or for the H2-A^b-restricted ISQAVHAAHAEINEAGR peptide derived from OVA_323–339, respectively, were also used in this study (11, 12). Mice were bred and maintained under specific pathogen-free conditions at the Central Animal Facility of Hannover Medical School or were purchased from Charles River Laboratories. All animal experiments were conducted in accordance with local and institutional guidelines.

Bone marrow cells from CX₃CR1^+/gfp mice were harvested by flushing femurs and tibias. Applying a cocktail of lineage-specific Abs (CD3, NK1.1, Ly6G, CD19, Ter-119), anti-CD11c, anti-c-kit mAb (CD117), and eGFP as a marker for CX₃CR1 expression, four populations of bone marrow cells were separated by flow sorting (see Fig. 1). The purity of sorted populations was always >96%. After cell sorting, the four populations were subsequently cultured in 24-well plates at 1–5 × 10⁵ cells/well in complete medium (RPMI 1640, 10% FCS, 2 mM l-glutamine, 1 mM sodium pyruvate, 10 mM HEPES, 50 μg/ml gentamicin, and 5 × 10⁻⁵ M 2-ME) supplemented with 30 ng/ml GM-CSF (PeproTech/Tebu). Every two to 3 days, the culture medium was collected, centrifuged to recover the cells in suspension, and then replated in fresh medium. Phenotypic analysis was performed by flow cytometry to monitor the course of DC differentiation. GFP⁺c-kit⁺ bone marrow cells cultured for 5 days under GM-CSF treatment were sorted into CD11c⁺CD103⁺ and CD11c⁺CD103⁻ DC (purity >99%). Cytospins were prepared, air-dried, fixed in 3% paraformaldehyde, and stained with H&E.

CD45.1 recipient mice were immunized i.p. with 150 μg of OVA (grade VI; Sigma-Aldrich) in 200 μl aluminum hydroxide gel adjuvant (2.0% Alhydrogel, Brenntag Biosector). Ten to 15 days later, 3 × 10⁵ lin⁻CX₃CR1⁺c-kit⁺ bone marrow cells were adoptively transferred to nonirradiated or irradiated recipients (6 Gy) that were treated in an aerosol chamber with OVA aerosol (1% in water) using a PariBoy vaporizer over 7 days as described previously (13). At day 8, mice were sacrificed and lung, spleen, and peripheral LN were analyzed by flow cytometry to monitor for the presence of CD45.2 donor-derived cells.

Twenty-four hours after the last aerosol treatment, mice were anesthetized and exsanguinated by perfusing the hearts with cold PBS until the lung lobes were free of blood. Lungs were perfused intratracheally three times with 1 ml each of Ca²⁺ - and Mg²⁺-free PBS supplemented with 0.1 mM EDTA (Invitrogen) and the BALF was collected. RBC were lysed using ammonium chloride lysis buffer (ACK, BioWhittaker). Lung tissue was then cut into small pieces and digested with collagenase A (Roche) plus 25 μg/ml DNase I (Roche) for 90 min at 37°C in a shaking platform (150 rpm). The digested mixture was passed through a nylon mesh to remove undigested fragments and then subjected to a Percoll gradient. Spleen or brLN donor-derived cells were obtained by enzymatic digestion using complete medium supplemented with DNase I and collagenase A or collagenase D (Roche), respectively, as previously described (9).

The following mAbs were used in this study: CD45.1 (A20), CD45.2 (104), CD3 (17A2), CD11b (M1/70.15), CD11c (HL3), CD19 (1D3), CD80 (16-10A1), CD86 (RMMP-1), CD40 (3/23), rat isotype control (R35-95), NK1.1 (PK136), Ter-119 (Ly-76), MHC class II (IA^b) (AF6-120.1), Ly6G (1A8), CD117 (104D2), and CD103 (M290) were purchased from BD Biosciences; PDL1 (MIH5), PDL2 (122), B7-H4 (188), 4-1BBL (TSK-1), and CD83 (Michel-17) were obtained from e-Bioscience. Fc receptors were blocked by preincubating the cell suspensions with 2 μg/ml of blocking anti-FcR mAb (2.4G2) before adding the specified mAb to reduce nonspecific binding (14). Dead cells and debris were excluded from acquisition by staining with DAPI (4′,6′-diamido-2-phenylindole hydrochloride) or propidium iodide (PI) staining. Flow cytometry acquisition was conducted on a LSR II cytometer (BD Biosciences) and data analysis was performed using WinList version 5 (Verity Software House).

Murine syngeneic B6, allogeneic BALB/c, OT-I Thy1.1, or OT-II Ly5.1 splenocytes were prepared and used as responder cells. To this end, spleens were harvested, minced under sterile conditions, and red cells were lysed in ACK lysing buffer for 2 min. Cells were then washed and suspended in complete medium. CD103⁺ and CD103⁻ DC were sorted from GFP⁺c-kit⁺ bone marrow cells that were differentiated in vitro in the presence of GM-CSF for 5 days, seeded in U-bottom 96-well plates (5000 DC/well), and cocultured with 1.5 × 10⁵ allogeneic BALB/c splenocytes or 1.5 × 10⁵ OT-I or OT-II T cells for 3 days. DC cultured with OT-I or OT-II cells were loaded with OVA prior culture. Proliferation of syngeneic, allogeneic, OT-I, and OT-II T cells in triplicate wells was then measured by pulsing the cells with 1 μCi/well methyl-[³H]thymidine for 16 h. Plates were harvested and thymidine uptake was quantified in a beta counter (MicroBeta TriLux, PerkinElmer).

Total RNA was extracted from CD103⁻ and CD103⁺ DC differentiated for 5 days from lin⁻CX₃CR1⁺c-kit⁺ bone marrow cells using RNeasy Mini kit (Qiagen), transcribed into Cy3- or Cy5-labeled cRNA, respectively, and cohybridized onto the same microarray. Samples derived from two independent experiments were analyzed separately on two arrays, including one dye-swap experiment. Whole Mouse Genome (4 × 44,000) Oligo Microarray kits (Agilent Technologies) were used in this study.

Sample processing, hybridization, and data extraction were performed according to standard protocols (Agilent Technologies). Detailed information concerning experimental procedures as well as all microarray data discussed in this publication have been deposited in the Gene National Center for Biotechnology Information Expression Omnibus (GEO, www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO series accession number GSE10882, or directly via the following link: www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token = dbmhvccekomusjy&acc = GSE10882. Criteria for visualization of microarray data for Figs. 3–6 are as follows: 1) All family members of a gene family of interest were selected as described in the legends of Figs. 3–6. 2) Data from family members that do not show significant expression were excluded. Significant mRNA expression was defined by algorithms implemented within the data extraction software: feature extraction V.9.1.3.1 (FEV9.1.3.1). Data were used for visualization when at least in one channel of each of the two replicate microarray experiments significant signal values were reached (corresponding to a “positive entry of 1” in the column labeled “IsWellAboveBackground” of the FEV9.1.3.1 results file). 3) When multiple probes directed against the same mRNA species were present on the microarray, data were visualized for that particular probe that showed highest intensity levels across all four channels of both microarrays (calculated by the arithmetic mean value of the “ProcessedSignal” values of all four subdatasets). 4) Highly significant differential mRNA expression was defined as follows: “PValue” as calculated by FEV9.1.3.1 for the corresponding probe was below 0.0001.

DC (2 × 10⁴) derived from GFP⁺c-kit⁺ bone marrow cells collected at day 5 of culture were incubated with 2.5 μg/ml OVA-Cy5 at 37°C or 4°C for 10 min, washed extensively in PBS, and analyzed by flow cytometry.

Gram-negative Escherichia coli TOP10 (Invitrogen) and zymosan A (Sigma-Aldrich) were labeled with the fluorescent dye Cy5 (Amersham). A bacterial cell suspension of OD equivalent to 1 measured at 600 nm was prepared, whereas zymosan A particles were resuspended at 1 mg/ml. Then, 100 and 300 μl of Cy5 were added to the bacterial and zymosan A suspensions, respectively. After 1 h of incubation in the dark, the reaction was stopped by adding cold PBS. The bacterial and zymosan A suspensions were centrifuged at 10,000 rpm for 5 min and washed several times to remove the unbound fluorochrome. Cy5-labeled bacteria and zymosan A were incubated with opsonizing polyclonal Ab (Molecular Probes) for 15 min and then washed and used in the phagocytosis assay. Phagocytosis assays were performed by incubating the nonopsonized or opsonized Cy5-labeled bacteria or Cy5-labeled zymosan A with GFP⁺c-kit⁺-derived DC (cultured for 5 days) and left for 20 min at 37°C. Cells were washed in cold PBS containing 0.02% Na₂-EDTA and samples were analyzed by flow cytometry.

CD103⁺ and CD103⁻ DC were sorted at day 5 from in vitro cultures of GFP⁺c-kit⁺ cells and incubated overnight in complete medium supplemented with either 1 μg/ml R848 (InvivoGen), 16 μg/ml CpG (Amersham), 1 μg/ml LPS (Sigma-Aldrich), or were left untreated. Supernatants were analyzed for the presence of IL-6, IL-10, MCP-1, IFN-γ, TNF, and IL-12 (p70 subunit) using a cytometric bead array following the manufacturer’s instructions (BD Biosciences).

Bone marrow cells of CX₃CR1^+/gfp heterozygous B6 mice were stained with a cocktail of lineage-specific mAb (CD3, NK1.1, Ly6G, CD19, Ter-119) excluding CD11b, together with anti-c-kit and anti-CD11c. Together with eGFP signals identifying CX₃CR1⁺ cells, several populations of lin⁻ cells could be identified (Fig. 1). Two CD11c⁺ subsets, one being GFP^low (R2) one being GFP^high (R3), as well as a GFP^intCD11c⁻ population (R4) were readily detectable. Further analysis of the latter subset revealed two populations of progenitor cells: GFP⁺c-kit⁻ (R5) and GFP⁺c-kit⁺ (R6). The latter population represents ∼20% of the GFP⁺ cells in bone marrow of CX₃CR1^gfp/+ mice. These GFP⁺c-kit⁺ cells are similar to the recently described macrophage and DC progenitor (MDP) progenitors (lin⁻CX₃CR1⁺c-kit⁺CD11b⁻) (15), although note that we did not include anti-CD11b mAb in the linage cocktail for reasons discussed in the Discussion.

Further analysis of the two CD11c⁺ subsets revealed that these cells represent immature stages of committed DC that do not further differentiate in vitro (data not shown). Thus, these populations were not further analyzed in the present study. To evaluate the potential of GFP⁺c-kit⁺ and of GFP⁺c-kit⁻ to give rise to CD11c⁺ DC following in vitro culture, both populations were sorted and incubated for 5 days in the presence of GM-CSF. Since c-kit is a surface marker of hematopoietic progenitor cells that is progressively down-regulated during cell differentiation (16), we analyzed c-kit expression of GFP⁺c-kit⁺ during a period of 7 days of in vitro culture. Starting at day 1, c-kit was continuously down-regulated, while as early as day 3, CD11c⁺ cells could be identified. At day 7, c-kit⁺ cells were virtually absent, while ∼70% of all cells expressed CD11c, suggesting that these cells differentiated to DC (Fig. 2,A). Next, we tested whether cells of GFP⁺c-kit⁺ or GFP⁺c-kit⁻ could be differentiated in vitro to CD103⁺ DC. As shown in Fig. 2,B, considerable amounts of CD11c⁺CD103⁺ and CD11c⁺CD103⁻ cells could be identified at day 5 in cultures of both GFP⁺c-kit⁺ as well as GFP⁺c-kit⁻ cells. A subset of cells expressing a low amount of CD103 was already present at day 3 in GFP⁺c-kit⁻ cell cultures (Fig. 2,B). Data obtained from these in vitro experiments indicate that GFP⁺c-kit⁺ as well as GFP⁺c-kit⁻ cells possess the capacity to differentiate to CD11c⁺CD103⁺ and CD11c⁺CD103⁻ DC. Cytospins of sorted CD11c⁺CD103⁺ and CD11c⁺CD103⁻ DC cells differentiated from GM-CSF-treated GFP⁺c-kit⁺ progenitors cultured for 5 days exhibited a DC-like morphology with a kidney-shaped nuclei and long cytoplasmatic extensions, as shown in Fig. 2 C.

We next compared the capacity of GFP⁺c-kit⁺ and GFP⁺c-kit⁻ cells to proliferate and expand in vitro in the presence of GM-CSF. c-kit⁻ bone marrow cells (GFP⁺c-kit⁻) survived without further expansion for 5 days but were progressively diminished at later time points. In contrast, c-kit⁺ cells (GFP⁺c-kit⁺) expanded 50-fold during the first 5 days of culture. Cell counts then stayed rather constant until day 10 and slowly decayed afterward (Fig. 2,D). To test whether CD103⁺ and CD103⁻ DC keep their phenotype regarding CD103 expression, CD11c⁺CD103⁻ and CD11c⁺CD103⁺ DC were sorted from day 5 cultures of GFP⁺c-kit⁺ cells and were further cultured for 2 days in the absence or presence of LPS (1 μg/ml). Interestingly, under both culture conditions, DC did not change their CD103 expression status while, as expected, LPS induced DC maturation as shown by increased expression of MHC class II molecules (Fig. 2 E and data not shown).

Based on our previous studies showing that CD103⁺ and CD103⁻ brLN DC differ with regard to their capacity to present and cross-present Ag (9), and based on the stable expression of CD103 observed in the above-described experiments, we reasoned that CD103 might also be a suitable marker that allows the differentiation of functionally distinct DC subsets generated in vitro. To test this hypothesis, we analyzed gene expression of CD11c⁺CD103⁺ and CD11c⁺CD103⁻ DC sorted from day 5 applying high-density oligonucleotide arrays. Data obtained from DC differentiated from GFP⁺c-kit⁺ cells cultures in the absence of maturation stimuli indicated that CD103⁻ DC express lower levels of several costimulatory molecules of the TNF and Ig superfamily, as well as MHC class II molecules (Fig. 3,A). Interestingly, CD103⁻ DC expressed higher levels of B7-H3 (CD276) (Fig. 3,A). Flow cytometry on resting DC, as well as on DC activated with LPS for 24 h, confirmed the tendency that CD103⁻ DC show reduced levels for some costimulatory molecules such as CD83, CD86, and PDL2, while others such as CD40, CD80, and PDL1 were expressed to similar amounts on CD103⁻ and CD103⁺ DC (Fig. 3 B).

To test whether CD103⁺ and CD103⁻ DC differ with regard to Ag presentation, both populations were used in allogeneic T cell proliferation assays. While both DC populations were equally potent to induce syngeneic T cell proliferation, CD103⁺ DC were ∼3-fold more efficient in stimulating allogeneic T cells compared with CD103⁻ DC (Fig. 3,C). However, analysis of T cells activated with specific Ag gave different results. Neither immature CD103⁺ nor immature CD103⁻ DC, loaded with OVA, were able to activate OVA-specific MHCI-restricted OT-I or MHCII-restricted OT-II cells (Fig. 3,D). Applying LPS-matured DC in this assay, we observed that CD103⁻ and CD103⁺ DC activated OT-I cells to a similar degree, while CD103⁻ DC were seemingly more efficient in activating OT-II cells (Fig. 3 D).

Macrophages and DC express a large repertoire of pattern recognition receptors (PRR) (17). Among the PRRs, TLRs, which regulate gene expression, are usually distinguished from those involved in effector immune functions such as phagocytosis (18, 19, 20). Within the scavenger receptor (SR)-A family, Msr1 and MARCO genes were 6.4- and 12.5-fold increased in CD103⁻ compared with CD103⁺ DC. Similarly, within the SR-B family CD36, Scarb1 and Scarb2 gene expression were increased 5.4-, 4.1-, and 2.2-fold, respectively. Genes from SR-D family (CD68) and SR-E (LOX-1) were up-regulated 3.8- and 7.9-fold, whereas SR-PSOX (CXCL16) was down-regulated 2.2-fold in CD103⁻ DC. There was also a significantly increased expression of other scavenger-like receptor genes in CD103⁻ compared with CD103⁺ DC such as Fpr-rs2 (formyl peptide receptor, 25.1-fold), CD93 (C1qr1, 17.9-fold), CD163 (hemoglobin scavenger receptor, 12.2-fold), or CD14 (2.6-fold), whereas the mRNA level for CD6 was 2.6-fold decreased (Fig. 4 A).

Three major groups of proteins within the C-type lectin family are known to play a role in the process of phagocytosis: group VI (mannose receptor family), group II (CD209, Langerin and Clec molecules), and group V (NK cell receptor-like molecules) (18). Remarkably, the mRNA level for these phagocytic receptors was highly up-regulated in CD103⁻ compared with CD103⁺ DC. Within group VI, the mannose receptor (Mrc1) was strongly up-regulated (73.1-fold), while the expression of DEC-205 was slightly diminished (−1.9-fold; Fig. 4,A). The high mRNA level for the mannose receptor correlated with an enhanced ability of the CD103⁻ DC to engulf Cy5-labeled OVA (Fig. 4 B), a process known to be mediated via this receptor (21).

Within the C-type lectin group II, mRNA expression levels of all CD209 members (CD209a–e) were moderately increased (between 1.6- and 3.4-fold), while Langerin was 4.9-fold increased in CD103⁺ DC. In contrast, members of the Clec family were increased 2- to 18-fold in the CD103⁻ DC (Fig. 4,A). Within the NK cell receptor-like molecules of C-type lectin group V, Clec7a (Dectin-1, β-glucan receptor) and Clec1b genes were 4.3- and 8.6-fold increased, while another set of genes within this group, such as CD94, CD72, and CD69, were 6.6-, 3.2-, and 4.8-fold decreased in CD103⁺ cells (Fig. 4 A).

Further groups of phagocytosis receptors studied were ficollins (Fcna and Fcnb) and chinitases (Chi3l3, Chi3l4, and Chi3l1). All mRNAs for these molecules were remarkably up-regulated in CD103⁻ DC, with Chi3l3 (52.2-fold) showing the strongest difference in expression compared with CD103⁺ DC (Fig. 4,A). The overall up-regulation of phagocytosis-associated genes in the CD103⁻ DC correlated with their capacity to engulf fluorochrome-labeled E. coli and yeast particles (zymosan) from S. cerevisiae (Fig. 4 C).

Optimal phagocytosis of some microorganisms depends on the deposition of serum opsonins on the surface of the microbe that coat the target cells and facilitate the uptake and subsequent destruction upon interaction with Fcγ and complement receptors. Most genes encoding for the Fc receptors were either below detection threshold (CD89, poly-IgR) or expressed at comparable levels (CD64, Fcer1a, and FcRN) in CD103⁻ and CD103⁺ DC (data not shown). Only Fcgr3a (CD16), Fcgr2b (CD32), and Fcer1g (high-affinity IgE receptor) were significantly higher expressed in CD103⁻ DC (Fig. 4,A). Regarding the genes encoding for the complement receptors CR2 (CD21), CR3 (CD11b/CD18), CR4 (CD11c/CD18), and C3ar1, comparable levels of mRNA expression were detected in both DC subsets (data not shown). Significantly higher expression levels of CD93 (C1qr1, 17.9-fold) and C5ar1 (4.6-fold) were found in CD103⁻ DC (Fig. 4 A and data not shown).

Microarray analysis also revealed increased transcripts for the genes encoding most of the TLRs (TLR1, TLR2, TLR4, TLR6, TLR7, and TLR8) in CD103⁻ DC, with TLR7 mRNA being in particular up-regulated (34-fold; Fig. 5,A). This observation prompted us to assess whether this differential pattern of TLR receptor expression was associated with a differential capacity to secrete proinflammatory cytokines when stimulated with TLR ligands. To that end, supernatants of stimulated CD103⁺ and CD103⁻ DC were quantified by cytometric bead arrays. Consistent with the higher levels of TLR4 and TLR7 expression on CD103⁻ DC, the addition of LPS (TLR4 agonist) and R848 (TLR7 agonist) induced a stronger production of TNF-α and IL-6 in stimulated CD103⁻ DC compared with CD103⁺ DC, while no significant changes of secretion of IL-10, MCP-1, IFN-γ, or IL-12 cytokines could be observed (Fig. 5,B). The TLR9 agonist CpG induced equivalent amounts of TNF-α in both populations but slightly less IL-6 secretion in CD103⁻ DC. This finding cannot be attributed to TLR9 mRNA expression levels, which were comparable in both DC cell populations (Fig. 5 B).

With regard to chemokine and chemokine receptor expression we observed that most chemokines that were differentially expressed were up-regulated in CD103⁻ DC, suggesting that these cells might recruit T cells (via CXCL10 and CXCL4), neutrophils (CXCL1, CXCL2), and monocytes to places of infection and inflammation. Chemokine receptors relevant for the recruitment of monocytes to sites of inflammation are CCR2 and CX₃CR1, although CCR1 and CCR5 play also a substantial role in this process (22). The detailed analysis of the microarray experiments suggests that CD103⁻ DC seemingly secrete higher amounts of chemokines than do the CD103⁺ DC, particularly those that would attract monocytes such as the ligands for CCR2 (CCL2, CCL6, CCL7, and CCL9). With regard to chemokine receptors, CCR6 and CCR7 were up-regulated in CD103⁺ DC, while CD103⁻ DC expressed higher levels of CCR2 (Fig. 6 A). Additional chemokine receptors, namely CCR1, CCR4, and CXCR4, were significantly detectable but do not show strong differences in mRNA abundance.

Apart from those mRNAs differentially expressed in CD103⁻ and CD103⁺ DC reported so far, many others were also differentially regulated between both subsets. Among them, signal regulatory protein β1 precursor (sirpβ1) and thrombospondin 1 (thbs1) were 103.9- and 39.4-fold up-regulated in CD103⁻ DC, respectively. It is known that sirpβ1 plays a role in regulating phagocytosis of CD8α⁻ (but not CD8α⁺) splenic DC (23), whereas thbs1 is mainly involved in apoptotic cell engulfment (24) (Fig. 6,B). CD103⁺ DC are known to build tight junction strands with epithelial cells. Thus, it was no surprise to see that tjp1 (zonula occludens 1 gene, ZO-1) and claudin-1 (cldn1) mRNAs were 10.4- and 3-fold increased, respectively, in CD103⁺ DC (Fig. 6 B and data not shown).

We next examined whether GFP⁺c-kit⁺ cells might contribute in vivo to CD103⁻ and CD103⁺ DC. To that end, GFP⁺c-kit⁺ bone marrow cells were adoptively transferred into nonirradiated or irradiated recipients in which a local lung inflammation had been induced by repetitive administration of OVA aerosol. Eight days after adoptive transfer of GFP⁺c-kit⁺ cells into sublethally irradiated CD45.1 recipients undergoing a local inflammatory response in the lung, DC derived from the adoptively transferred donor cells could be identified in spleen, lung, brLN, and mesenteric LN. Of interest, 50–60% of these DC express CD103 (Fig. 7,A). With regard to absolute numbers, donor-derived DC were preferentially found in spleen and brLN (Fig. 7 B). We failed to identify donor-derived cells expressing CD19, CD3, or NK1.1 cells, indicating that the adoptively transferred GFP⁺c-kit⁺ cells predominantly differentiate to DC under the experimental conditions chosen. Taken together, these in vivo observations support the hypothesis that GFP⁺c-kit⁺ bone marrow cells can differentiate to CD103⁺ and CD103⁻ DC capable of migrating to inflamed tissues and their draining LN.

CD103 (α_E) is the α-chain of the α_Eβ₇ integrin that is expressed on human and mouse lymphocytes and facilitates the adhesion of these cells to epithelia via its binding to E-cadherin (25, 26). Intestinal CD11b⁺CD103⁺ DC possess the ability to induce CCR9⁺α₄β₇⁺ gut-tropic CD8⁺ effector T cells (7). Most CD103⁺ DC in the spleen also expresses CD8α (M.-L. del Rio and R. Förster, unpublished results), a DC subset known to be localized in the T cell area in this organ. CD8α⁺ DC, however, are almost absent in the lung, where CD11c⁺CD11b⁺ and plasmacytoid DC (CD11c^lowGr1⁺B220⁺120G8⁺) are the predominant DC populations (27). We and others have previously shown that within the CD11c⁺CD11b⁺ DC, the expression of CD103 allows to distinguish two subpopulations: CD103⁺ (CD11c^highCD11b^low) DC, which represents the major DC subset in the lung, and a CD103⁻ (CD11c^int CD11b^high) DC subset (8, 9).

Two subsets of bone marrow progenitor cells giving rise to defined myeloid lineages have recently been identified in CX₃CR1^+/gfp mice: myeloblast progenitor (lin⁻c-kit⁺CX₃CR1⁻) and MDP (lin⁻c-kit⁺CX₃CR1⁺) (15). MDP have been identified in particular in the bone marrow of CX₃CR1^+/gfp mice (lin⁻CX₃CR1⁺c-kit⁺) as the main progenitor intermediary for monocytes and macrophages, as well as DC (both CD11b⁺ and CD11b⁻) (15), but not for plasmacytoid DCs. Plasmacytoid DCs develop from a distinct bone marrow precursor cell differentiated in vitro in the presence of Flt3L (28, 29). Fogg et al. have shown that MDP immediately convert in vitro into CD11c⁺CD11b⁺ DC cells upon exposure to GM-CSF, but not into macrophages, which were only derived after culturing MDP with M-CSF in vitro (15). Our study made use of “MDP-like” progenitors (lin⁻CX₃CR1⁺c-kit⁺CD11b⁺/CD11b⁻), which are similar to the MDP progenitor described by Fogg et al. with the only difference that in our work, CD11b was not used in the lineage cocktail to ensure that c-kit⁺CX₃CR1⁺CD11b⁻ (MDP) and their immediate progenies (c-kit⁺CX₃CR1⁺CD11b⁺) were represented in the progenitor cell population (15). This consideration was based on the fact that within the lung, both CD103⁺ and CD103⁻ DC express CD11b. Thus, to assess whether GFP⁺c-kit⁺ and GFP⁺c-kit⁻ bone marrow cells are able to generate CD103⁺ and CD103⁻ DC in vitro, these progenitors were sorted and differentiated in vitro in the presence of GM-CSF. Although both populations were able to give rise to CD11c⁺CD103⁺ and CD11c⁺CD103⁻ DC, we focused on the GFP⁺c-kit⁺ cells, which, compared with GFP⁺c-kit⁻ cells, possess a high ability to proliferate and expand in vitro.

Data obtained in the present study support a model in which CD103⁺ and CD103⁻ DC generated from GFP⁺c-kit⁺ cells represent two distinct DC populations: no substantial conversion from sorted CD103⁻ DC into CD103⁺ DC (or vice versa) was observed within the following 4 days of culture (Fig. 2 E and data not shown), arguing against the idea that one of the two populations serves as a progenitor for the other. Furthermore, since sorted CD103⁺ and CD103⁻ DC kept their CD103 expression profile even in the presence of LPS, a TLR4 ligand known to induce DC maturation, we also trust that expression of CD103 is not a maturation marker. Both nonmatured CD103⁺ and CD103⁻ DC displayed low expression of MHCII as well as costimulatory molecules, but they showed significant functional differences with regard to many aspects of DC biology, suggesting that these two DC subsets might have acquired marked specialization required to face distinct requirements.

CD103⁻ DC showed a higher capacity than CD103⁺ DC to engulf nonopsonized as well as opsonized microorganisms and to secrete proinflammatory cytokines upon ligation with TLR agonists. Similar to our findings, CD103⁻ DC isolated from mesenteric LN have been reported to secrete proinflammatory cytokines (mainly TNF-α and IL-6) following stimulation with TLR agonists (30).

The relationship between macrophages and DC is still a subject of intense debate (31). It remains unclear which array of molecules would allow discrimination between DC and macrophages (32). Indeed, CD103⁻ DC display several markers that are also found on macrophages and also possess some features that could be attributed to macrophages (e.g., high phagocytic activity, secretion of proinflammatory cytokines upon ligation with TLR, poor costimulatory activity of allogeneic T cells). Nevertheless, we provide evidence that the CD103⁻ cells that are derived from GFP⁺c-kit⁺ bone marrow cells cultured in the continuous presence of GM-CSF are indeed bona fide DC. As described in the present study for GFP⁺c-kit⁺ cells, MDP precursors grown in the presence of GM-CSF but in the absence of stromal cells give rise to CD11c⁺CD11b⁺ DC but not to macrophages. Furthermore, these cells, as the cell described in the present study, display a DC-like morphology and phenotype (15). Finally, the CD103⁻ cells of the present study express CD11c, a marker widely used for the identification of DC in mice.

Chemokines are essential components of the innate as well as the adaptive immune system (33, 34). Within the lung, CD103⁻ DC produce higher levels of chemokines than CD103⁺ DC or even macrophages, making them a potential target in asthma therapy (35, 36). In vitro-generated CD103⁻ DC are also considerably more efficient than CD103⁺ DC in secreting a large variety of inflammatory chemokines. Thus, it is tempting to speculate that once a first wave of monocytes/progenitors reaches the inflamed tissue and differentiates to DC, CD103⁻ DC derived from these precursors would initiate the secretion of chemokines that then would not only attract neutrophils but also monocytes through CCR1, CCR2, and CCR5 (22). This mechanism would act as a positive feedback loop that ensures the maintenance of a continuous flow of DC precursors indispensable for the maintenance of the DC pool and for the persistence of an inflammatory immune response.

The CD103⁺ and CD103⁻ DC subsets generated in vitro from GFP⁺c-kit⁺ bone marrow cells did not display the capacity to differentially stimulate CD8 and CD4 Ag-specific T cells that we recently reported for lung-derived CD103⁺ and CD103⁻ DC that had taken up Ag in vivo in the lung (9). The reason for this difference is currently unclear. However, it seems likely that DC residing in peripheral organs such as the lung receive additional signals that further drive their differentiation leading to distinct Ag presentation and cross-presentation profile. Since CD103⁺ DC reside between epithelial cells, it seems possible that these epithelial cells provide defined differentiation factors that are not present in our in vitro cultures.

The mechanisms that are involved in replenishing pulmonary DC are not well understood, but it has recently been suggested that monocytes contribute to the pool of pulmonary DC under both steady-state and inflammatory conditions (4). It is well established that lung DC, probably like most DC residing in peripheral organs, are subjected to an active process of cell turnover (37). Lung DC numbers decay rapidly after irradiation treatment due to their constant migration to the lung draining LN even in the absence of inflammation (2). This suggests that the maintenance of DC numbers in the lung requires the constant input of self-renewing cells or the continuous fill up by DC precursors.

It has been recently suggested that, under steady-state conditions, Ly6C^high and Ly6C^low monocytes from peripheral blood can differentiate into lung DC (4, 38). Interestingly, upon inflammation, Ly6C^high cells still give rise to DC while Ly6C^low cells primarily differentiate to macrophages and to a lesser extent to DC (38). Further studies done by Randolph and colleagues suggest that Ly6C^high and Ly6C^low cells differentiate to CD103⁺ DC and CD103⁻ CD11b^high lung DC, respectively (5). Additional studies revealed that Ly6C^high cells can give rise to DC and Ly6C^low cells to macrophages in infectious peritonitis and the healing myocardium, respectively (39, 40). Along this line, the adoptive transfer of GFP⁺c-kit⁺ progenitors described herein allows for an efficient repopulation of CD11c⁺CD103⁺ and CD11c⁺CD103⁻ DC in the lung of irradiated recipients undergoing allergic lung inflammation. Note that we cannot formally rule out that some of the donor-derived CD11c⁺CD103⁻ cells in the lung might be macrophages since lung macrophages also express CD11c and might have up-regulated CD11b due to the allergic airway inflammation (4, 8).

CD103⁺ DC are localized in the airway mucosa next to the basal surface of bronchial epithelial cells where E-cadherin is present, and they are therefore more readily exposed to airway-carrying stimuli than CD103⁻ DC, which are situated in the subepithelial regions (8). Based on their location, it seems possible that the differential distribution of both subpopulations and the refractory behavior of CD103⁺ DC to produce cytokines and chemokines may serve as a mechanism to prevent continuous leukocyte recruitment and their activation in response to slight perturbations of the lung microenvironment. In this model, CD103⁺ DC would be the first cell type of the innate immune system to sense the danger signal, while the CD103⁻ would be the innate mechanism responsible for the antigenic clearance that would activate an inflammatory response facilitating the recruitment of neutrophils and monocytes to the site of inflammation.

In summary, we show that CD103⁺ and CD103⁻ DC can be generated in vitro under GM-CSF treatment from GFP⁺c-kit⁺ and GFP⁺c-kit⁻ bone marrow cells. Compared with CD103⁺ DC, CD103⁻ DC exhibited enhanced phagocytosis activity and secretion of inflammatory cytokines. Furthermore, adoptive transfer of GFP⁺c-kit⁺ cells into aerosol-sensitized and subsequently irradiated mice is able to reconstitute CD103⁺ and CD103⁻ DC populations in lymphoid and nonlymphoid compartments, which indicates that these precursors contribute to the pool of CD103⁺ and CD103⁻ DC in the lung.

We thank Christina Reimer for cell sorting.

The authors have no financial conflicts of interest.