Complement receptor C5aR1/CD88 and dipeptidyl peptidase-4/CD26 define distinct hematopoietic lineages of dendritic cells

Abstract

Differential display of the integrins CD103 and CD11b are widely used to distinguish two major dendritic cell (DC) subsets in nonlymphoid tissues. CD103⁺ DCs arise from fms-like tyrosine kinase receptor 3 (FLT3)-dependent DC precursors (preDC), whereas CD11b^hi DCs can arise either from preDCs or FLT3-independent monocytes. Functional characterization of these two lineages of CD11b^hi DCs has been hindered by the lack of a widely applicable method to distinguish between them. We performed gene expression analysis of fractionated lung DCs from C57BL/6 mice, and found that monocyte-derived (mo)DCs including CD11b^hiLy-6C^lo tissue resident and CD11b^hiLy-6C^hi inflammatory moDCs express the complement 5a receptor 1 (C5aR1)/CD88, whereas preDC-derived conventional DCs (cDCs) including CD103⁺ and CD11b^hi cDCs express dipeptidyl peptidase-4/CD26. Flow cytometric analysis of multiple organs, including the kidney, liver, lung, lymph nodes, small intestine and spleen, confirmed that reciprocal display of CD88 and CD26 can reliably distinguish FLT3-independent moDCs from FLT3-dependent cDCs in C57BL/6 mice. Similar results were obtained when DCs from BALB/c mice were analyzed. Using this novel approach to study DCs in mediastinal lymph nodes, we observed that the majority of blood-derived LN resident DCs, as well as tissue-derived migratory DCs, are cDCs. Furthermore, cDCs, but not moDCs, stimulated naïve T cell proliferation. We anticipate that the use of antibodies against CD88 and CD26 to distinguish moDCs and cDCs in multiple organs and mouse strains will facilitate studies aimed at assigning specific functions to distinct DC lineages in immune responses.

Introduction

Dendritic cells (DCs) induce adaptive immunity by acquiring antigens and presenting peptides derived from them to naïve T cells (1). In nonlymphoid tissues such as the lung, two major CD11c⁺ DC subsets can be identified based on their display of the integrins alpha E (CD103) and alpha M (CD11b) (2, 3). Lung CD103⁺ DCs are a homogeneous population and are similar to CD8α⁺ DCs, which are primarily found in lymphoid tissues. Both of these DC types are derived exclusively from Fms-like tyrosine kinase receptor 3 (FLT3)-dependent preDCs and are therefore termed conventional DCs (cDCs) (3–7). By contrast, CD11b^hi DCs in the lung are a heterogeneous population of cells comprising both cDCs and monocyte-derived (mo)DCs. The latter include short-lived Ly-6C^hi inflammatory DCs as well as longer-lived Ly-6C^lo resident moDCs that are seen both in the steady state and during inflammation (3, 4, 8–11).

Considerable progress has been recently made towards understanding the molecular basis of DC development. In particular, several transcription factors have been found to have important roles in the progression of progenitor cells to mature DCs. CD103⁺ cDC development requires several such factors, including inhibitor of DNA binding (Id)2, interferon regulatory factor (IRF)8, basic leucine zipper ATF-like transcription factor (BATF)3 and nuclear factor interleukin 3 regulated (NFIL3) (4). A different set of transcription factors, including IRF4, contributes to CD11b^hi cDC development, although mice lacking these proteins retain substantial numbers of CD11b^hi DCs in some tissues. In contrast to transcription factors that affect either CD103⁺ cDCs or CD11b^hi cDCs, ZBTB46 is a transcription factor found in both types of cDCs (12). Mice expressing a Zbtb46 promoter-driven green fluorescent protein (GFP) or diphtheria toxin receptor allow ready identification and purification of all cDCs, at least during steady state conditions (12, 13). The lectin DNGR-1/CLEC9A is also produced in common DC precursor cells (CDPs), allowing fate mapping of their progeny, which include CD8α⁺ and CD103⁺ cDCs, as well as some CD11b^hi cDCs (14). Although these novel reporter mouse strains have improved our ability to identify cDCs, it is often necessary to discriminate between CD11b^hi cDCs and CD11b^hi moDCs in commonly used strains that do not carry reporter genes. For such experiments, antibodies must be used to discriminate among the various DC subsets, and considerable progress has also been made in this area. Antibodies against CD14 and CD64 are widely used to identify macrophages, and recent studies have shown that these molecules are also displayed by some CD11c⁺MHC-II^hiCD11b^hi DCs, presumably CD11b^hi moDCs (15). Accordingly, display of CD14 and CD64 has been used to identify moDCs, together with the marker CD24, which is displayed on some CD11b^hi cDCs (16, 17). This is a useful strategy for discriminating between cDCs and moDC in the lung and small intestine, but is of limited utility for distinguishing DC lineages in other organs (15, 17). For example, in skin-draining lymph nodes, cDCs identified by their high expression of Zbtb46 have relatively high levels of CD14 (12), which is associated with moDCs in the lung (11, 15). It is therefore important to develop novel strategies that can reliably discriminate between cDCs and moDCs in multiple organs and in multiple mouse strains. In the present study, we analyzed gene expression of moDCs and cDCs prepared from the lung to identify novel markers of these two lineages. We found that complement C5a receptor 1 (C5aR1)/CD88 is selectively expressed on moDCs, including Ly-6C^lo resident moDCs and Ly-6C^hi inflammatory moDCs. Conversely, CD103⁺ cDCs and CD11b^hi cDCs display high cell surface levels of dipeptidyl peptidase-4 (DPP4)/CD26. Antibodies against these molecules can reliably identify the major lineages of cDCs and moDCs in multiple peripheral and lymphoid tissues, including kidney, liver, lung, small intestine, spleen and lung-draining LNs.

Materials and Methods

Mice

BALB/c/J, C57BL/6J, CD45.1 (B6.SJL-Ptprc^a Pepc^b/BoyJ) and OT-II (B6.Cg-Tg(TcraTcrb)425Cbn/J) were purchased from Jackson Laboratories (Bar Harbor, ME). Flt3L^−/− (C57BL/6-Flt3L^tm1/mx) mice and control C57BL/6N mice were from Taconic (Germantown, NY). Mice were housed in specific pathogen-free conditions at the NIEHS, and used between 6 and 12 weeks of age in accordance with guidelines provided by the Institutional Animal Care and Use Committees.

Allergic sensitization

In some experiments, mice were anesthetized by isoflurane inhalation and given 50 µl PBS containing 0.1 µg LPS (Sigma, St. Louis, MO) together with 100 µg endotoxin-free OVA (Profos AG, Germany) (LPS/OVA), or 10 µl house dust extract (HDE) together with OVA (HDE/OVA) by oropharyngeal aspiration as described previously (18, 19).

Preparation and ex vivo analysis of DCs

DCs were prepared from mouse lung, mediastinal LNs (mLNs), spleen, kidney, liver, and from the small intestine. After removal of Peyer’s patches, small intestines were washed extensively in PBS containing 5mM EDTA, 145µg DTT and 10% FBS as previously reported (20). Minced tissues were digested as described previously (21), with minor modifications in incubation times as follows; 10 min for liver, 30 min for mLN, spleen, kidney and small intestine; and 60 min for lung. Low density cells, collected from gradient centrifugation using 16% Nycodenz (Accurate Chemical, Westbury, NY), were diluted to 1–2 × 10⁶/100 µl and incubated with a non-specific binding blocking reagent cocktail of anti-mouse CD16/CD32 (2.4G2), normal mouse and rat serum (Jackson ImmunoResearch, West Grove, PA) for 5 min. For staining of surface antigens, cells were incubated with fluorochrome [Allophycocyanin (APC), APC-Cy7, Alexa Fluor 488, Alexa Fluor 647, Brilliant Violet 510, Brilliant Ultra Violet 395, eFluor 450, eFluor 605 NC, FITC, PerCP-Cy5.5 or Phycoerythrin] or biotin-conjugated antibodies against mouse CD3ε (145-2C11), CD8α (53–6.7), CD11b (M1/70), CD11c (N418 and HL3), CD14 (Sa2–8), CD19 (605), CD24 (30-F1), CD26 (H194–112), CD45.1 (A20), CD45R–B220 (RA3–6B2), CD49b (DX5), CD64 (X54–5/7.1), CD88 (20/70), CD103 (M290), CD115 (AFS98), CD135 (A2F10), CD172a (P84), Ly-6A/E (D7), Ly-6C (AL-21), Ly-6G (1A8), Ly6-C/G (RB6–8C5), MHC class II I-A^b (AFb.120), I-A/I-E (M5/114), TER119 (TER-119) and Vβ5 TCR (MR9-4) (BD Biosciences (San Jose, CA), BioLegend (San Diego, CA) and eBioscience (San Diego, CA)). Staining with biotinylated antibodies was followed by fluorochrome-conjugated streptavidin. Stained cells were analyzed on a five laser LSR-II, or sorted on a five laser ARIA-II flow cytometer (BD Biosciences), and the data analyzed using FACS Diva (BD Bioscience) and FlowJo software (Treestar, Ashland, OR). For gated CD11b^hi DCs, we analyzed from 5 × 10³ to 2 × 10⁴ cells. Only single cells were analyzed, and dead cells were excluded based on their forward and side scatter.

Purification of monocytes and preDC, and their differentiation in vitro and in vivo

Bone marrow (BM) was collected, and red blood cells lysed with 0.15 M ammonium chloride; 1 mM potassium bicarbonate. Inflammatory monocytes (CD3⁻CD11b⁺CD11c⁻CD19⁻CD49b⁻CD115⁺F4/80⁺I-A⁻Ly-6A/E⁻Ly-6C^hiLy-6G⁻TER119⁻7-AAD⁻) or preDCs (B220⁻CD3⁻CD11b⁻CD11c⁺CD19⁻CD49b⁻CD135⁺CD172a^intI-A⁻Ly-6A/E⁻Ly-6G⁻TER119⁻) from BM were enriched using an automated magnet-activated cell sorter (AutoMACS) (Miltenyi, Auburn, CA) then further purified by flow cytometry-based sorting. For in vitro differentiation, BM cells, purified monocytes or preDCs were cultured in complete RPMI1640 containing 10% fetal bovine serum (FBS) (RPMI-10) and 100 ng/ml recombinant human FLT3L (generated in the Protein Expression Core, NIEHS) or 5 ng/ml recombinant mouse GM-CSF (R&D Systems, Minneapolis, MN) for up to 8 days (11). Cultured cells were harvested and analyzed by flow cytometry. In some experiments, cells were centrifuged onto glass slides using a Cytospin 4 centrifuge (Thermo, Waltham, MA) stained with hematoxylin and eosin, and analyzed on a microscope (Olympus, Center Valley, PA). To track monocyte-derived cells in vivo, C57BL/6 (CD45.2⁺) mice were given LPS/OVA by oropharyngeal aspiration, and 2 hours later received 7×10⁵ purified monocytes from CD45.1⁺ mice by tail vein injection. CD45.1⁺ donor cell-derived cells in recipient mice were subsequently analyzed by flow cytometry at the indicated time points.

DC culture

DCs were purified by flow cytometric sorting from lungs or mLNs of untreated mice or mice given HDE/OVA through the airway 16 h prior to tissue procurement. Purified DCs (1×10⁵) were cultured for 24 h in 200 µl complete RPMI-10 in a 5% CO₂ incubator, and IL-6 and TNF-α in the supernatants analyzed by enzyme-linked immunosorbent assay. To test their T cell-stimulating ability, purified DCs (6.25×10³/well) were cultured together with naive CD4⁺ T cells (CD8α⁻CD8β⁻CD11b⁻CD11c⁻CD19⁻CD44^loB220⁻CD49b⁻I-A⁻Ly-6C/G⁻TER119⁻) (5×10⁴/well) prepared from LNs and spleens of OT-II TCR transgenic mice in 200 µl Iscove’s modified Dulbecco’s medium containing 10% FBS (Certified, Life Technologies). To evaluate T cell proliferation, carboxyfluorescein succinimidyl ester (CFSE) (Life Technologies)-labeled CD4⁺ T cells were analyzed by flow cytometry after 3 days of culture, and viable T cells were counted under a microscope after 5 days.

Calcium flux

Intracellular calcium was measured using Fluo-4 (Life Technologies, Grand Island, NY) as described previously (22). Briefly, DCs in Hank’s balanced salt solution containing 1mM Ca²⁺, 1mM Mg⁺ and 0.5% BSA were labeled with 4 µM Fluo-4 for 30 min at 37 °C in the presence of 4 mM probenecid. After washing, cells were resuspended in 500 µl buffer, and allowed to rest for at least 15 min prior to analysis. The Fluo-4 signals were recorded by flow cytometry for 30 seconds prior to adding 10 µM C5a agonistic peptide (FKP-(D-Cha)-Cha-R) (AnaSpec, Fremont, CA), then monitored for 12 min. Ionomycin (2 µl of 1 mg/ml, EMD Millipore, Billerica, MA) was added at the end to ensure the presence of the calcium indicator.

Gene expression analysis

Microarray experiments were done in triplicate, with DCs purified from pooled lungs of 20 mice (total of 60 mice). The ranges of analyzed DC numbers were as follows: CD103⁺ DCs, 1.98–2.53×10⁵ cells; CD11b^hiCD14^loLy-6C^lo DCs, 3.04–3.54×10⁵ cells; CD11b^hiCD14^hiLy-6C^lo DCs, 1.24–1.37×10⁵ cells; CD11b^hiLy-6C^hi DCs, 3.35–8.11×10⁵ cells. Total RNA was isolated from each subset using RNeasy kit (Qiagen, Valencia, CA), and comprehensive gene expression analysis was conducted using Affymetrix Mouse genome 430 2.0 GeneChip arrays (Affymetrix). After scanning arrays, data was obtained using the GeneChip Command Console software using the MS5 algorithm (Affymetrix). To identify subset-specific genes, data were further analyzed using IPA software (Ingenuity). For quantitative PCR (qPCR) analysis, total RNA from purified lung DC subsets prepared independently from microarray was isolated using Trizol (Life Technologies). The total RNA was converted to cDNA with oligo dT primers and SuperScript III First Strand kit (Life Technologies). PCR amplification was performed with SYBR Green Master Mix (Applied Biosystems) and Mx3000P QPCR system (Agilent Technologies, Santa Clara, CA) using the following forward (F) and reverse (R) primers; Gapdh F: AACTTTGGCATTGTGGAAGG, R: AACTTTGGCATTGTGGAAGG; Il12a F: CTAGACAAGGGCATGCTGGT, R: GCTTCTCCCACAGGAGGTTT); Irf4 F: AATCCCCATTGAGCCAAGCA, R: TCGTCGTGGTCAGCTCTTTC; Irf8 F: ACAATCAGGAGGTGGATGCTT, R: CGTGGCTGGTTCAGCTTTGT; Nos2 F: CCCCGCTACTACTCCATCAG, R: GGCTTCAGGTTCCTGATCCAA; Tnf F: GATCGGTCCCCAAAGGGATGA, R: TGCTCCTCCACTTGGTGGTTT and Zbtb46 F: ATCACTTCTCACTACCGGCAT, R: AAGACGTTCTTATGTGCCTTGAA. The relative expression of each gene was normalized to Gapdh expression.

Statistics

Data are presented as mean ± SEM. Statistical differences between groups were calculated using a two-tailed Student t-test, unless indicated otherwise. P < 0.05 was considered significant.

Results

Monocyte-derived pulmonary CD11b^hi DCs selectively express C5aR1/CD88

Reciprocal display of CD64 and CD24 has been reported to define moDCs and cDCs, respectively, at least for DCs prepared from the lung or small intestine (15–17, 23). Using C57BL/6 mice, we confirmed that lung CD11b^hi DCs, defined as CD11b^hiCD11c⁺MHC class II⁺ autofluorescence^lo cells, (Supplementary Fig. 1A) can be clearly resolved into CD14^hi(CD64^hi)CD24^lo moDCs and CD14^lo(CD64^lo)CD24^hi cDCs (Supplementary Fig. 1B). Because BALB/c mice are frequently used in animal models of allergic diseases such as asthma, we also tested whether these same markers can similarly distinguish CD11b^hi cDCs and moDCs in that strain. Unexpectedly, antibodies against CD14, CD64 and CD24 failed to resolve CD11b^hi lung DCs from BALB/c mice into the distinct moDC and cDC populations. This was because a relatively small number of BALB/c DCs stained for CD24, and because many CD24^lo DCs were also CD14^lo or CD64^lo (Supplementary Fig. 1B). We next analyzed DCs from lung-draining mLNs of C57BL/6 and BALB/c mice and found that these antibodies also failed to clearly resolve CD11b^hi DCs into their cDC and moDC components. Unlike CD11b^hi DCs from C57BL/6 mouse lungs, CD24 display on LN DCs was not inversely correlated with display of either CD14 or CD64 (Supplementary Fig. 1C, 1D). The latter findings are consistent with the previously reported high levels of CD14 on CD11b^hi cDCs, and of ZBTB46-GFP on CD24^lo DCs in skin-draining LNs (11–13). Thus, neither CD14 nor CD64 can identify moDCs in all organs and mouse strains, underscoring the need for more widely applicable cell surface markers.

We reasoned that comprehensive gene expression analysis of well-defined populations of CD11b^hi moDC and cDC populations might reveal genes that are reciprocally expressed in these DC lineages, regardless of their anatomical location. Because display levels of CD14 are sufficient to identify CD11b^hi moDCs and cDCs in lungs of C57BL/6 mice, we purified CD11b^hiCD14^hiLy-6C^lo resident moDCs and CD11b^hiCD14^loLy-6C^lo cDCs from lungs of these animals using flow cytometry-based cell sorting (Fig. 1A). RNA was separately prepared from these two populations, and microarray-based gene expression analysis performed. As expected, expression of Cd14 was several-fold higher in moDCs than in cDCs, confirming effective separation of these two populations. We also found that three other cell membrane protein-encoding genes, C5ar1, C3ar1, Ptger2, were expressed at much higher levels in moDCs than in cDCs (Table 1). We focused on C5ar1, because it had the greatest difference in expression between moDCs and cDCs. This gene encodes the well-characterized complement receptor C5aR1 (CD88), and antibodies that recognize it are readily available. Flow cytometric analysis confirmed at the protein level that CD88 is displayed at higher levels on CD11b^hiCD14^hi and CD11b^hiLy-6C^hi moDCs than on either CD11b^hiCD14^lo cDCs or CD103⁺ cDCs (Fig. 1B–D). This was true during steady state conditions, and after oropharyngeal instillation of LPS/OVA, a procedure used for allergic sensitization through the airway (18, 24).

FIGURE 1.

Display of CD88/C5aR1 on lung moDCs. (A) Pre- and post-purification of CD11b^hiCD14^hiLy-6C^lo moDCs and CD11b^hiCD14^loLy-6C^lo cDCs from C57BL/6 mouse lungs following inhalation of LPS/OVA. Data shown are from a single experiment, representative of three. (B – D) Flow cytometric analyses of CD88 display on lung DC subsets from C57BL/6 mice. Similar results were obtained in 3 independent experiments. (B) CD88 display on CD11b^hiCD14^hi, CD11b^hiCD14^lo and CD103⁺ lung DCs at steady state. Mean fluorescence intensity (MFI) is indicated. (C) CD88 display on CD103⁺ DCs, CD11b^hiCD14^hi, CD11b^hiCD14^lo and CD11b^hiLy-6C^hi DCs after instillation of LPS/OVA. (D) MFI of CD88 display on lung DC subsets. Mean values ± SEM from 3 mice are shown. * P<0.01, ** P<0.001 and *** P<0.0001. (E) Co-display of CD88 and CD64 on CD11b^hi lung DCs. (F) C5a–mediated Ca²⁺ influx in the indicated lung DCs purified from lungs of LPS/OVA-treated C57BL/6 mice. DCs were loaded with Fluo-4, and fluorescence measured by flow cytometry before and after treatment with C5a peptide and ionomycin. 8 to 14 ×10⁴ cells per sample were acquired, and 909 ± 295 cells were analyzed for each time point. *P<0.05, Fluo-4 signals in C5a–treated cells compared to mock-treated cells. A representative result from three independent experiments is shown.

Table 1.

Genes encoding cell membrane proteins highly expressed by CD11b^hiCD14^hi lung DCs

Sequence name	CD14hi/CD14lo fold change	ANOVA p-value	CD14hi DC intensity	CD14lo DC intensity
C5ar1	32	0.00155	1537	47
C3ar1	20	0.00101	1505	72
Ptger2	7	0.00023	98	13
Cd14	5	0.00063	4162	700

Cell membrane protein-encoding genes expressed >5-fold higher in CD14^hi DCs than in CD14^lo DCs in microarray analysis. Mean intensity, ANOVA p-value from 3 replicates (20 mice each) and fold changes are indicated.

Because CD64 has been used previously to identify moDCs (15, 16), we studied this marker and found that it was displayed on CD88^hi CD11b^hi lung DCs (Fig. 1E). Another macrophage marker, MerTK (15), was also displayed on CD88^hi CD11b^hi DCs, at least during steady state (Supplementary Figure 1E). However, following LPS/OVA instillation to induce inflammation, lung CD11b^hiCD88^hi moDCs comprised 2 populations; MerTK^positive Ly-6C^lo lung resident moDCs and MerTK^low/negative Ly-6C^hi inflammatory moDCs. Thus, MerTK could be used to identify mature, lung resident moDCs, but not inflammatory DCs.

We further tested whether surface display of CD88 correlates with responsiveness to C5a, using the Ca²⁺-dependent fluorescent indicator, Fluo-4 (22). Using CD64 as a moDC marker to avoid neutralization of C5aR by antibodies against CD88, we found that C5a can induce Ca²⁺ influx in resident moDCs (CD11b^hiCD64^hi) and in inflammatory moDCs (CD11b^hiLy-6C^hi), but not in CD103⁺ or CD11b^hiCD64^lo cDCs (Fig. 1F). These results indicate that CD88 is selectively displayed on the surface of cells having the characteristics of moDCs and that it is functional in those cells.

Conventional pulmonary CD11b^hi DCs selectively express DPP4/CD26

We next sought to identify a novel cell surface marker that is specific for cDCs. We compared gene expression in two cDC subsets (CD103⁺ DCs and CD11b^hiCD14^lo DCs) with that of two moDC subsets (CD11b^hiCD14^hi DCs and CD11b^hiLy-6C^hi DCs) (Supplementary Fig. 2A–C). As expected from previous studies, many genes (including Xcr1) were differentially expressed between CD103⁺ and CD11b^hi cDCs (25–27). Of greater relevance to the goal at hand, we also identified fifty-two genes that were expressed at least 2-fold higher in CD11b^hi cDCs than in CD11b^hi moDCs, and eighteen of those genes encoded membrane proteins (Figure 2A, Supplementary Fig. 2C, Table 2). The gene encoding dipeptidyl peptidase-4 (DPP4)/CD26 had the highest expression in CD11b^hiCD14^lo cDCs (Table 2). Flow cytometric analyses confirmed that CD103⁺ cDCs uniformly display high levels of CD26 at the protein level, whereas CD11b^hi DCs in the lung comprise two major populations that can be distinguished by their relative display of CD26. These two populations were observed at steady state, as well as during LPS/OVA-induced inflammation (Fig. 2B). Display of CD26 inversely correlated with that of CD88, suggesting that CD26 is highly displayed by cDCs, but not moDCs.

FIGURE 2.

CD26/DPP4 display on lung cDCs. (A) Venn diagram indicating the cellular location of proteins encoded by genes expressed >2-fold higher in CD103⁺ cDCs and in CD11b^hi cDCs than in either moDC subset. (B) Display of CD26 on CD103⁺ and CD11b^hi lung DCs during steady state conditions and after instillation of LPS/OVA. (C) Mean fluorescent intensity (MFI) ± SEM of CD26 display on lung DC subsets. Data are from 2 mice. * P<0.01, ** P<0.001 and N.S. not significant (P>0.05).

Table 2.

Intensities of genes encoding plasma membrane proteins selectively expressed by cDCs

Sequence name	CD11bhi CD14lo	CD103+	CD11bhi CD14hi	CD11bhi Ly-6Chi
Dpp4	1329	1709	486	131
H2-Ob	1187	1215	219	115
Btla	1148	2518	387	125
Grasp	1137	2261	356	443
P2ry10	1136	1870	493	153
Flt3	904	1940	256	32
Fyn	884	963	363	175
Tnfrsf18	549	1163	247	74
Htr7	431	1105	74	112
Gnb4	350	2076	142	62
Chst15	335	493	105	26
Cldn1	271	2392	117	9
Epcam	252	785	85	25
Tmem8	237	410	80	92
Tnfrsf4	230	1044	51	13
Kit	224	600	36	26
Cdon	86	322	42	17
Prnp	78	275	31	21

Mean intensity in microarray analysis from 3 replicates (20 mice each) is shown.

CD11b^hiCD88^hi DCs are monocyte-derived

To confirm that CD11b^hiCD88^hiCD26^lo DCs are indeed moDCs, we purified Ly-6C^hi monocytes from CD45.1 mouse BM (Supplementary Fig. 3A, 3B), adoptively transferred them into CD45.2 mice and instilled LPS/OVA into the airways of recipient mice to induce inflammation. One day after this treatment, many CD45.1⁺ donor-derived monocytes had differentiated into inflammatory DCs, as evidenced by their acquisition of CD11c and MHC-II. Freshly isolated monocytes did not display CD88 (Fig. 3A), but acquired its expression upon their differentiation to DCs (Fig. 3B). By day 3 post-transfer, the majority of donor-derived cells no longer displayed high levels of Ly-6C, but maintained display of CD88 (Fig. 3B). Very few CD88^lo donor cells were seen at either time point, confirming that monocytes give rise almost exclusively to CD88^hi moDCs.

FIGURE 3.

Confirmation of developmental lineages of CD88^hi and CD26^hi lung DCs. (A) CD88 display on monocytes freshly isolated from BM. (B) Differentiation of CD45.1 monocytes following their adoptive transfer into CD45.2 recipient mice. Shown are the gating strategy to identify CD45.1⁺ donor cells, as well as their display of CD88 and Ly-6C one or three days after instillation of LPS/OVA into recipient mice. (C–E) Flow cytometric analysis of lung DCs of wild-type and Flt3L^−/− mice at steady state and 16 hours after instillation of LPS/OVA. (C) Cytograms and percentages of CD11b^hiCD88^loCD26^hi and CD11b^hiCD88^hiCD26^lo lung DCs. (D) Percentages of CD103⁺ lung DCs. (E) Cytograms and percentages of CD11b^hiCD88^loLy-6C^lo, CD11b^hiCD88^hiLy-6C^lo, CD11b^hiCD88^hiLy-6C^hi lung DCs. (n=3). *P<0.05. Results shown are from one of three experiments yielding similar results.

To confirm that CD88^loCD26^hi cells are cDCs, we studied mice lacking FLT3L, a cytokine that promotes cDC development. In agreement with previous reports (8, 11), Flt3L^−/− mice were essentially devoid of CD103⁺ cDCs, and had severe reductions in CD11b^hiCD88^loCD26^hi DCs in the lung, (Fig. 3C, 3D). By contrast, CD11b^hiCD88^hi DCs were abundant in Flt3L^−/− mice (Fig. 3C, 3E), as would be expected for cells derived from monocytes. Thus, multiple independent experiments confirmed that at least for lungs of C57BL/6 mice, CD88 and CD26 are reliable markers to identify moDCs and cDCs, respectively.

Our finding that CD11b^hi DCs of BALB/c mice are not effectively resolved into distinct populations by antibodies against CD64 and CD14 (Supplementary Fig. 1) prompted us to test if CD26 and CD88 are more effective in this regard. We harvested DCs from the lung of BALB/c mice and found that the latter two molecules could indeed resolve CD11b⁺ DCs into two distinct populations: CD88^hiCD26^lo DCs and CD26^hiCD88^lo DCs (Supplementary Fig. 4A). Using this approach to evaluate DC changes in the lung following LPS/OVA instillation, we observed that cell numbers for all types of CD11b^hi DCs increased, with the largest increase seen for CD88^hiCD26^loLy-6C^hi inflammatory DCs.

Analysis of moDCs and cDCs in the lung

To further characterize moDCs and cDCs in the lung, we used flow cytometry-based cell sorting to purify CD103⁺ cDCs, CD11b^hiCD88^loCD26^hi cDCs, and CD11b^hiCD88^hiCD26^lo moDCs from untreated mice and from mice that had inhaled LPS/OVA. Analysis of gene expression by qPCR revealed that CD11b^hiCD88^hiCD26^lo moDCs highly expressed Il12a, Tnf, and Nos2, which are highly expressed in monocyte-derived cells, such as inflammatory DCs and inflammatory macrophages (28, 29). These genes were not highly expressed in CD103⁺ cDCs or in CD11b^hiCD88^loCD26^hi cDCs (Fig. 4). Conversely, CD103⁺ cDCs and CD11b^hiCD88^loCD26^hi cDCs exclusively expressed Zbtb46, which is expressed by all cDCs (12, 13). As expected, Irf8 was expressed in CD103⁺ cDCs, whereas Irf4 was expressed in CD11b^hiCD88^loCD26^hi cDCs (Fig. 4).

FIGURE 4.

Gene expression in lung DC subsets. mRNA for the indicated genes in different DC subsets following their purification from pooled lungs of 10 C57BL/6 mice at steady state or 16 hours after instillation of LPS/OVA. Duplicate samples were assayed by qPCR and values normalized to Gapdh. N.D., not detectable. * P<0.05, ** P<0.01, *** P<0.001. Results shown are from one of two experiments yielding similar results.

Analysis of moDCs and cDCs in lung-draining LNs

During steady state, the majority of DCs in skin-draining and mensenteric LNs are tissue-derived CD11c^int MHC-II^hi migratory DCs and blood-derived CD11c^hi MHC-II^int resident DCs. Both of these DC populations express Zbtb46, suggesting they are cDCs (12, 13). However, during inflammation, many blood-derived moDCs are recruited to the LN (4, 30, 31), and these cells also display high levels of MHC class II (32). Thus, when analyzing cells from inflamed LNs, high levels of MHCII cannot distinguish migratory cDCs from moDCs. We therefore tested whether these two DC populations can be distinguished using antibodies against CD88 and CD26. During steady state, CD11b^hiCD88^loCD26^hi cDCs were the major DC population, followed by CD103⁺ cDCs, and relatively few CD11b^hiCD88^hiCD26^lo DCs were seen (Fig. 5A–C). A marked reduction of CD11b^hiCD88^loCD26^hi DCs in Flt3L^−/− mouse LNs confirmed that these cells are indeed cDCs (Fig. 5B, C). Following instillation of LPS/OVA to induce mild inflammation in the lung, the number of CD103⁺ cDCs in mLNs increased dramatically and these cells became the major DC population. Modest increases were seen for CD11b^hiCD88^loLy-6C^lo(CD26^hi) cDCs, CD11b^hiCD88^hiLy-6C^lo(CD26^lo) moDCs and CD11b^hiCD88^hiLy-6C^hi(CD26^lo) inflammatory moDCs in both C57BL/6 and BALB/c mice (Fig. 5C, Supplementary Fig. 4B), but the two moDC subsets remained minor populations. Importantly, the strict dependence of CD11b^hiCD88^loCD26^hi DCs on the presence of FLT3L regardless of immunological status indicates that CD88 and CD26 can reliably resolve two distinct lineages of DCs in LNs.

FIGURE 5.

Analysis of DCs in mLNs. (A, B) Flow cytometric analysis of CD11b and CD103 on total DCs (A), and CD88, CD26 and Ly-6C on CD11b^hi DCs (B) from mLN of wild-type and Flt3L^−/− mice at steady state and after instillation of LPS/OVA. (C) Cell numbers of the indicated mLN DC subsets in wild-type and Flt3L^−/− mice at steady state and after instillation of LPS/OVA. Mean values ± SEM are shown (n=3). *P<0.05 for comparing wild-type and Flt3L^−/− mice. Results shown are from one of two experiments yielding similar results.

T cell activation by moDCs and cDCs

An important characteristic of DCs is their ability to stimulate proliferation of naïve T cells. To evaluate this property in moDCs and cDCs prepared from the lung, we cultured naïve CD4⁺ T cells from OVA-specific TCR (OT-II) transgenic mice together with various lung DC subsets. Because house dust extracts (HDE) represent an environmentally relevant source of adjuvants promoting allergic airway inflammation (19, 22), HDE/OVA was administered to the airways of mice prior to preparing DCs from their lungs. Both cDC subsets (CD103⁺ cDCs and CD11b^hiCD88^lo cDCs) induced robust naïve CD4 T cell proliferation as measured by dilution of CFSE, and each purified subset was more efficient in this regard than were total DCs (Fig. 6A, 6B). By contrast, neither moDC subset (CD11b^hiCD88^hi resident moDCs or CD11b^hiLy-6C^hi inflammatory moDCs) induced proliferation of CD4⁺ T cells. T cell counts following the co-cultures confirmed that both cDC subsets stimulated T cell proliferation much better than did moDCs (Fig. 6C). These results demonstrate that, at least in the lung, moDCs do not efficiently stimulate naïve T cell proliferation, suggesting they fulfill a different function, such as production of proinflammatory cytokines. In support of this, moDCs produced much larger amounts of IL-6 and TNF-α than did cDCs (Fig. 6D).

FIGURE 6.

Activation of naïve CD4 T cells and cytokine production by DCs. (A–C) Lung DCs were purified from C57BL/6 mice by flow cytometry 16 h after instillation of HDE/OVA, and cultured with naïve CD4 T cells prepared from OT-II mice at 1:8 ratio (6.25×10³ DCs and 5×10⁴ T cells). (A, B) CFSE-labeled naïve CD4 T cells were cultured with the indicated lung DC subsets. Three days later, CFSE fluorescence in CD4⁺Vβ5TCR⁺ T cells was analyzed by flow cytometry. Representative histograms of CFSE levels in T cells (A) and percentage of CFSE^low proliferating T cells from triplicate assay (B) are shown. (C) Viable T cells were counted after 5 days of culture, and fold-expansion over initial input of T cells was calculated. (D) IL-6 and TNF-α in culture supernatants of the lung DC subsets, as measured by ELISA. *P<0.05, **P<0.01, ***P< 0.001. Results shown are from one of at least two experiments yielding similar results. (E–I) mLN DCs were purified from C57BL/6 mice by flow cytometry 1 day (24 hours) (E–G) or 3 days (H, I) after instillation of HDE/OVA and cultured with OT-II naïve CD4 T cells at 1:8 ratio (6.25×10³ DCs and 5×10⁴ T cells). (E, F, H) CFSE-labeled naïve CD4 T cells were cultured with the indicated mLN DC subsets. Three days later, CFSE fluorescence in CD4⁺Vβ5TCR⁺ T cells were analyzed by flow cytometry. Representative histograms of CFSE levels in T cells (E) and percentage of CFSE^low proliferating T cells from duplicate assays (F, H) are shown. (G) Viable T cells were counted after 5 days co-culture of T cells with DCs prepared 1 day after HDE/OVA instillation (n=2). (I) Viable T cells were counted after 5 days of co-culture of T cells with DCs prepared 3 days after HDE/OVA instillation (n=3–4). * P<0.05, **p<0.01, *** P< 0.001. Results shown are from one of at least two experiments yielding similar results.

LNs that drain peripheral tissue contain both resident DCs and migratory DCs; the former develop from precursors arriving to LNs directly from the blood, whereas the latter migrate from peripheral tissue to LNs through the lymphatics (4, 31). Compared to resident DCs, migratory DCs display higher amounts of MHC class II I-A, but lower levels of CD11c (20). We found that both types of DCs were markedly reduced in mLNs of Flt3L^−/− mice (Supplementary Figure 3C), in agreement with our previous results (Fig. 5) and indicating that the majority of resident DCs, as well as migratory DCs, are cDCs. Although resident DCs in skin-draining LNs acquire soluble antigens that are passively carried to LNs through the lymphatics and stimulate T cells (31, 33, 34), it has been unclear whether resident DCs in lung-draining mLN are also capable of stimulating naïve T cells, and if this is dependent on their developmental lineage. To address this issue, we prepared total mLN DCs, and separately purified CD103⁺ migratory cDCs, CD11b^hiCD88^loI-A^hiCD11c^int migratory cDCs, CD11b^hiCD88^loI-A^intCD11c^hi resident cDCs and CD11b^hiCD88^hiLy-6C^hi inflammatory moDCs from mLNs of mice one day after instillation of HDE/OVA to the airway. Total mLN DCs, as well as both types of migratory cDCs, induced robust proliferation of naïve CD4⁺ T cells as measured by CFSE-dilution assay and by cell count, whereas mLN resident cDCs and inflammatory moDCs did not (Fig. 6E–G). To test the possibility that a longer time is required for resident DCs to acquire T cell stimulating activity, we cultured naïve CD4⁺ T cells with DCs prepared from mLN three days after allergic sensitization. Although CD103⁺ DCs harvested at this time point displayed a reduced ability to stimulate naïve CD4⁺ T cell proliferation compared to their counterparts harvested at one day post sensitization, both migratory cDC subsets – CD103⁺ DCs and CD11b^hiCD88b^loI-A^hi – still robustly stimulated T cell proliferation, whereas resident cDCs and moDCs did not (Fig. 6H and 6I). These results suggest that at least in the allergic sensitization model tested here, the ability to stimulate naïve CD4⁺ T cell proliferation is restricted to the lung-derived migratory cDCs.

Use of CD88 to identify in vitro-generated moDCs

BM is a rich source of DC progenitors and is therefore widely used to generate large numbers of DCs for study. Typically, BM cells are cultured in media containing either FLT3L or GM-CSF. We found that after culture in FLT3L-containing media, approximately 90% of BM cell-derived DCs (BMDCs) were CD88^lo, suggesting that these conditions primarily give rise to cDCs. By contrast, BM cultured with GM-CSF gave rise to both CD88^lo and CD88^hi DCs (Supplementary Fig. 3D). Including IL-4 in the media together with GM-CSF, a procedure widely used for generation of BMDCs, failed to yield large numbers of CD88^hi cells (data not shown), likely because IL-4 suppresses Cd88 mRNA expression in these moDCs (35). To determine whether precursor cell identity determines CD88 display on their DC progeny, we separately purified monocytes and preDCs from BM and cultured them with GM-CSF without IL-4. Freshly isolated monocytes did not display CD88 (Fig. 3A), but the majority of monocyte-derived CD11c⁺MHC-II⁺ cells after culture were CD88^hi (Supplementary Fig. 3E). By contrast, CD11c⁺MHC-II⁺ cells derived from preDCs were primarily CD88^lo. These data provide additional evidence that CD88 display on DCs is determined by their developmental lineage, and further suggest that CD88 display can be used to resolve BMDCs into their moDC and cDC components.

CD88 and CD26 distinguish moDCs and cDCs in multiple organs

Previous studies have shown that although CD24 and CD64 can be used to distinguish CD11b^hi moDCs from CD11b^hi cDCs in the lung and small intestine, these markers do not discriminate between lineages of DCs prepared from other organs (17). We confirmed that splenic CD8α⁺ DCs, as well as CD103⁺ DCs in kidney, liver and small intestine display a CD64^loCD24^hi phenotype, but that CD11b^hi DCs in the spleen, kidney and liver are not resolved into their moDC and cDC components (Supplementary Fig. 4C). We therefore tested the ability of antibodies to CD88 and CD26 to distinguish cDCs from moDCs in various tissues. CD8α⁺ DCs and CD11b^hi DCs prepared from the spleen all displayed a CD88^loCD26^hi cDC phenotype (Fig. 7A), consistent with previous reports that there are very few moDCs in the spleen (9, 12). This was confirmed by marked reductions of CD11b^hiCD26^hi DCs and CD8α⁺ DCs in spleens of Flt3L^−/− mice (Fig. 7B). In the kidney, liver and small intestine, CD103⁺ DCs were uniformly CD88^loCD26^hi. Importantly, CD11b^hi DCs were effectively resolved into CD88^loCD26^hi DCs and CD88^hiCD26^lo DCs (Fig. 7A and 7B). The former were Flt3L–dependent, whereas the latter were Flt3L-independent, confirming their identities as cDCs and moDCs, respectively. These results demonstrate that reciprocal display of CD88 and CD26 is widely useful to resolve two distinct lineages of DCs in many different organs.

FIGURE 7.

Display of CD88 and CD26 on DCs from multiple organs. (A) CD88 and CD26 display on CD11b^hi and CD8α⁺ DCs from spleen, and on CD11b^hi and CD103⁺ DCs from the kidney, liver and small intestine of untreated C57BL/6 mice. (B) Cell number of CD8α⁺, CD103⁺ and CD11b^hiCD88^hiCD26^lo and CD11b^hiCD88^loCD26^hi DC subsets in indicated organs of untreated wild-type and Flt3L^−/− mice. *P<0.05. Results shown are from one of two experiments yielding similar results.

Discussion

Because of their ability to link the innate and adaptive arms of the immune response, DCs are currently the focus of intense investigation. In recent years, it has become clear that there are many types of DCs, and that they can be subdivided based on their developmental lineage or display of different cell surface markers. However, assignment of specific functions to these different DC subsets has been difficult, in part because cell surface markers that can unequivocally resolve CD11b^hi moDCs from their CD11b^hi cDC counterparts in multiple organs and multiple strains have been lacking. In the current study, we found that CD88 is highly displayed on the surface of moDCs and that this display can distinguish them from cDCs. Although we initially identified CD88 as a marker of moDCs prepared from lungs, subsequent experiments revealed that it is also expressed in moDCs from several other organs including the kidney, liver and small intestine. For these organs, CD88^hi DCs were present at normal numbers in Flt3L^−/− mice, which display marked reductions in cDCs, but not moDCs. Additionally, adoptive transfer of monocytes gave rise almost exclusively to CD88^hi cells, and CD88 was highly displayed on moDCs in multiple organs of two different mouse strains. These observations are in agreement with a study of C5aR-GFP reporter mice, which showed that C5aR-GFP fluorescence associates with monocyte/macrophage markers such as CD11b, F4/80 or Ly-6C/G, but not the cDC marker CD8α or the plasmacytoid DC marker PDCA-1 (36). We did not conduct an exhaustive study of all organs, however, and it is possible that some of them contain DC populations whose lineages cannot be identified solely by display of CD26 and CD88.

Although our current work focused on reciprocal display of CD88 and CD26 as a means to define moDC and cDC populations, respectively, it is possible that these molecules also affect DC function. C5a elicited Ca²⁺ influx in moDCs, indicating that CD88-mediated signaling is functional in these cells. Previous studies have shown Cd88^−/− mice have exacerbated allergic inflammation upon allergen inhalation (37–39), and that Cd88^−/− pulmonary DCs promote increased production of chemokines by T cells (37). However, in vitro generated Cd88^−/− BMDCs produce lower levels of Th17-inducing cytokines, such as IL-1β, IL-23 and TGF-β, than do wild-type BMDCs, and have a diminished capability to induce allergic inflammation upon adoptive transfer (40). Thus, the role of CD88 in DC function remains unresolved.

The role of CD26 in DC function is also poorly understood. This transmembrane glycoprotein has many functions, including peptidase activity, T cell co-stimulatory activity and cell adhesion properties (41, 42). Although Cd26^−/− mice display enhanced allergic airway inflammation (43), it is unclear whether this is due to altered function of DCs or lymphocytes, as CD26 is also expressed by T and B cells (42). Our observation that CD26 is highly expressed by cDCs is consistent with some previous reports (17, 23), and the ability to selectively delete CD26 as well as CD88 on DCs should be helpful to understand the role of these molecules in DC function.

We found that in the spleen, most CD11b^hi DCs (as well as CD8α⁺ DCs) are CD88^loCD26^hi. This suggests that most splenic DCs are cDCs, in agreement with previous reports (9, 12). However, the relative abundance of moDCs and cDCs in most tissue-draining LNs is still unclear, although skin-draining LNs contain both Zbtb46-dependent and independent cells (44). We previously found that although cDCs in the lung can migrate to regional mLNs, moDCs do not (11). Consequently, lung-derived DCs in regional LNs are almost exclusively cDCs. In addition to these migratory DCs, LNs also contain non-migratory (LN resident) DCs that probably arise from precursors that have direct access to LNs from the blood. The developmental lineage of these LN resident DCs has been controversial (4, 12, 30), despite the fact that they are the major population of DCs in LNs, at least during steady state conditions. Our ability to discriminate between these lineages by their reciprocal display of CD88 and CD26 prompted us to study their abundance in mLNs. We observed that the majority of mLN resident DCs (CD11b^hiI-A^intCD11c^hi), as well as the majority of migratory DCs (CD103⁺ and CD11b^hiI-A^hiCD11c^int), are cDCs. This is consistent with the finding that although blood-derived monocytes can accumulate in LN in the setting of severe inflammation (11, 31), this does not happen to an appreciable extent during steady state conditions, or even following allergic sensitization (11). It seems likely, therefore, that most LN resident DCs develop in situ from preDCs, which have been previously shown to reside in lymphoid organs, including LNs (45).

An important characteristic of DCs is their ability to stimulate naïve T cells. We found that lung-derived migratory CD103⁺ cDCs and migratory CD11b^hi cDCs induced robust proliferation of naïve CD4⁺ T cells, whereas moDCs did not. This finding is in agreement with a recent study showing that CD64⁺ moDCs poorly stimulate naïve T cells (17), and consistent with a role of moDCs as suppressive cells. However, some groups have reported that moDCs can promote T cell proliferation (32, 46). It remains to be seen whether some of the apparently discordant results of previous studies are due to differences in DCs from different LNs, or from incomplete separation of moDCs from cDCs. We anticipate that the strategy we have developed to clearly resolve moDCs from cDCs in multiple tissues will be helpful in this regard.

Our finding that mLN resident CD11b^hi cDCs poorly stimulated T cell proliferation was somewhat unexpected because in skin-draining LNs, resident cDCs are reported to activate T cells and induce their proliferation in vivo (33, 47). It is possible that antigens drain more efficiently to skin LNs than to mLNs because although antigen was readily detected in resident DCs of skin-draining LNs after immunization with OVA or MHC I-Eα protein (31, 33), we were unable to detect fluorescent OVA in mLN resident DCs even after instilling high amounts of OVA (Nakano et al, unpublished observations). Alternatively, LN resident DCs in skin-draining LNs might possess different activities than their counterparts in mLNs, possibly because they are exposed to different environmental factors. It is also conceivable that migratory DCs and resident DCs cooperate to stimulate T cells in mLNs, as they are reported to do in skin-draining LNs (34). Regardless of the correct explanation, the ability to discriminate each LN resident DC subset using reliable markers should facilitate future studies of their development and functions in immune responses.

The definition of DCs based on their morphology, surface molecule expression and function has recently been called into question because macrophages and DCs can share some of these features (48). CD11c^hiMHC-II⁺ cells are widely regarded to be DCs, but some macrophages can also display high amounts of these cell surface proteins (11, 49, 50). Similarly, some activated macrophages exhibit dendrite-like pseudopodia usually associated with DCs (51). Even the ability to activate naïve T cells, one of the functional definitions of DCs, can be observed for some macrophages (50). It has therefore been proposed that classification of DCs should be based on their hematopoietic lineage with the term ‘DC’ being reserved for cDCs (17, 52, 53). According to this proposed definition, neither inflammatory moDCs, nor tissue resident moDCs, would be classified as DCs. The most widely used approach to generate human DCs is to culture blood-derived monocytes with appropriate stimuli, such as GM-CSF and IL-4. Similar approaches are used to generate mouse DCs, which possess all known features of DCs (32), but these cells would no longer be called DCs according to the proposed definition. It remains to be seen whether this new definition will be widely adopted, but it might help to standardize nomenclature and therefore communication of findings. Regardless of how moDCs are defined, it had been difficult to reliably determine their ontogeny in animals that have not been genetically altered. Our current finding that reciprocal display of CD88 and CD26 can distinguish cDCs (or DCs) from moDCs (or monocyte-derived cells) in multiple organs and in different mouse strains should facilitate studies of how these different cell types initiate, propagate and regulate immune responses.

Abbreviations used

C5aR1

Complement 5a receptor 1

CCR

CC chemokine receptor

cDCs

conventional DCs

CDP

common DC precursor

CFSE

carboxyfluorescein succinimidyl ester

DPP4

dipeptidyl peptidase-4

FLT3

Fms-like tyrosine kinase 3

FLT3L

FLT3 ligand

HDE

house dust extract

LPS

lipopolysaccharide

mLNs

mediastinal lymph nodes

moDCs

monocyte-derived DCs

preDC

DC precursor

qPCR

quantitative PCR