Human Blood CD1c⁺ Dendritic Cells Encompass CD5^high and CD5^low Subsets That Differ Significantly in Phenotype, Gene Expression, and Functions

Dendritic cells (DCs) are professional APCs that control the initiation, orientation, and magnitude of immune responses (1–3). DCs exhibit heterogeneity with respect to surface marker expression, anatomic localization, and physiological function. Both human and mouse DCs comprise two major subsets: plasmacytoid DCs (pDCs) and conventional DCs (cDCs). cDCs can be further divided into IFN regulatory factor (IRF)8–dependent cDC1s and IRF4-dependent cDC2s (4–6). In mice, cDC1s are more potent in cross-presenting Ags to CD8⁺ T cells, whereas cDC2s are superior in priming CD4⁺ T cells. The similar division of labor between DC subsets had also been proposed for human cDC1s and cDC2s (7–10). However, human cDC2s isolated from the lymphoid organ could efficiently cross-present Ag to CD8⁺ T cells (11, 12). The identification and characterization of functionally distinct DC subpopulations is important for understanding the mechanisms of immune regulation and for optimizing DC-based immune therapies.

Beyond the division of labor between DC subsets, there are developmentally and functionally different subpopulations within the mouse cDC2 subset (3, 13–15). The mouse spleen contains two cDC2 subpopulations according to the expression level of endothelial cell–specific adhesion molecule (Esam). The development of Esam^high cDC2s depends on Notch2, but the development of Esam^low cDC2s occurs independently of Notch2 (15). Notch2 is also essential for the development of cDC2s in the intestinal lamina propria that induce CD4⁺ Th cells expressing IL-17 (Th17) (15). Conversely, cDC2s that develop in response to the transcription factor KLF4 promote Th2 rather than Th17 polarization (13).

DCs can also be classified into lymphoid tissue DCs or non–lymphoid tissue DCs based on their anatomical localizations. Non–lymphoid tissue DCs charged with tissue Ags can traffic from peripheral tissue to lymph nodes (LNs) under the control of CCR7 and its ligand CCL21 and CCL19, which are referred to as migratory DCs (16–18). In the steady-state, migratory DCs are superior to LN-resident DCs in inducing self-antigen–specific regulatory T cells to maintain peripheral tolerance (19–22). It has been reported that IRF4 is important not only for cDC2 development but also for its migration from peripheral tissue to draining LNs in mice (23–26). In human LNs, migratory DCs of both cDC1 and cDC2 subsets have been identified (12, 27).

Until recently, functionally distinct human cDC2 subpopulations have not been identified. In the present study, we showed that human blood cDC2s can divided into two distinct subpopulations according to CD5 expression. We show that CD5^high DCs and CD5^low DCs differ significantly in gene expression, cytokine production, Ag presentation, and T cell polarization.

PBMCs from healthy volunteers were isolated via density gradient centrifugation using Ficoll-Paque Plus (17-1440-02; GE Healthcare). Tonsils were obtained from consenting patients undergoing routine tonsillectomy for recurrent tonsillitis. No patient was suffering from tonsillitis at the time of operation. LNs were obtained from cancer patients undergoing surgery. All samples were obtained in accordance with institutional ethical guidelines. The samples were cut into small fragments and digested with 0.5 mg/ml collagenase D (Roche) in the presence of 5 μg/ml DNase (Roche) for 30 min. Afterward, the cells were filtered with a 70-μm cell strainer (Corning). Then, the cells were washed twice with PBS containing 2% FBS and 2 mM EDTA (Amresco). Nonspecific binding was blocked using Fc receptor–blocking solution (BioLegend). Subsequently, the cells were stained with the following Abs at 4°C: anti-lineage (CD3, CD14, CD16, CD19 and CD20)–FITC, anti–CD5-Brilliant Violet 421, anti–CD123-PerCP-Cy5.5, anti–CD11c-allophycocyanin, anti–HLA-DR-allophycocyanin-Cy7, and anti–CD1c-PE-Cy7. The lineage (Lin)⁺ cells were gated out, and the surface molecule expression profiles of HLA-DR⁺CD11c⁺CD1c⁺CD5^high and HLA-DR⁺CD11c⁺CD1c⁺CD5^low DCs were analyzed. When detecting intracellular molecules, the PBMCs were permeabilized with Perm/Wash buffer (BD Biosciences) according to the manufacturer’s protocol. All flow cytometry was performed using an LSRFortessa cell analyzer (BD Biosciences), and the data were analyzed with Summit 4.3 software (Dako, Glostrup, Denmark). Detailed information about the Abs used is presented in Supplemental Table I.

PBMCs were labeled with purified mouse anti-human mAbs against Ags, including CD3, CD14, CD16, and CD19 in PBS supplemented with 2% FBS and 2 mM EDTA for 30 min at 4°C. After washing, the cells were incubated with goat anti-mouse IgG microbeads (Miltenyi Biotec) for 30 min at 4°C. Then, labeled cells were magnetically depleted using an LD column, which was placed in the magnetic field of a MidiMACS separator (Miltenyi Biotec). The enriched fractions were stained with the following Abs at 4°C: anti–Lin-FITC, anti–CD5-PE, anti–CD1c-PE-Cy7, anti–CD123-PerCP-Cy5.5, anti–CD11c-allophycocyanin, and anti–HLA-DR-allophycocyanin-Cy7. Lin⁻HLA-DR⁺CD123⁻CD11c⁺CD1c⁺ CD5^high and CD5^low DCs were sorted using a BD FACSAria cell sorter (BD Biosciences). Both CD5^high and CD5^low CD1c⁺ DCs were isolated to >95% purity using these procedures. The selected cells were maintained in complete medium consisting of RPMI 1640 medium (Life Technologies) supplemented with 10% FBS (Life Technologies), 2 mM l-glutamine (Invitrogen), 100 U/ml penicillin/streptomycin (Invitrogen), 0.1 mM MEM–nonessential amino acid solution (Life Technologies), and 1 mM sodium pyruvate (Invitrogen) at 37°C in an incubator with an air atmosphere containing 5% CO₂.

Purified CD5^high and CD5^low DCs were centrifuged and labeled with Wright–Giemsa stain. Images were captured using a Nikon DS-Ri2 microscope.

Total RNA was extracted from fresh or stimulated cells with TRIzol (Invitrogen) according to the manufacturer’s instructions. Oligo(dT) primers and Moloney murine leukemia virus reverse transcriptase (M1705; Promega) were used for cDNA synthesis. All of the gene transcripts were quantified by real-time PCR using SYBR Green quantitative PCR master mix (S7563; Applied Biosciences/Life Technologies) and a Rotor-Gene 65H0 (Corbett Life Science) according to the manufacturer’s instructions: initial denaturation at 95°C for 10 min followed by 40 cycles at 95°C for 15 s and at 60°C for 60 s. Elongation factor 1-α (EF1-α) was used as a housekeeping control gene to normalize the amount of cDNA among samples. The specific primers for RT-PCR were: EF1-α, 5′-ATATGGTTCCTGGCAAGCCC-3′, 5′-GTGGGGTGGCAGGTATTAGG-3′; TLR2, 5′-GTGCCCATTGCTCTTTCAC-3′, 5′-CATTGGATGTCAGCACCAGA-3′; TLR3, 5′-AAAACCTTTGCCTTCTGCACG-3′, 5′-TTCCAGCTGAACCTGAGTTCC-3′; TLR7, 5′-CATGCTCTGCTCTCTTCAACC-3′, 5′-TGGGGAGATGTCTGGTATG-TG-3′; TLR8, 5′-GCTGCTGCAAGTTACGGAATG-3′, 5′-CTTCGGCGCATAA-CTCACAG-3′; TLR9, 5′-GAAGGGACCTCGAGTGTGAA-3′, 5′-GTGCTGCCATGGAGAAGTG-3′; MAFB, 5′-ACTTGAGCGAGAGGGAGGAA-3′, 5′-CCTTGGTGACTTCTCGGGAC-3′; IRF4, 5′-TCCGAGAAGGCATCGACAAG-3′, 5′-AGGCGTTGTCATGGTGTAGG-3′; CD5, 5′-ATGCAGCCTGACAACTCCTC-3′, 5′-ACAAACAGGTCTGGCTCTCG-3′; IFN-γ, 5′-GGCTTTTCAGCTCTGCATCG-3′, 5′-TCTGTCACTCTCCTCTTTCCA-3′; IL-22, 5′-ATGGGGACCCTGGCCACCAG-3′, 5′-AGCCAGCATGAAGGTGCGGT-3′; IL-10, 5′-AAGACCCAGACATCAAGGCG-3′, 5′-ATCGATGACAGCGCCGTAG-3′; IL-4, 5′-TCCGATTCCTGAAACGGCTC-3′, 5′-CGTACTCTGGTTGGCTTCC-3′; IL-17A, 5′-CCGCCACTTGGGCTGCATCA-3′, 5′-AGCCGGAAGGAGTTGGGGCA-3′; IFN-β, 5′-GTCTCCTCCAAATTGCTCTC-3′, 5′-ACAGGAGCTTCTGACACTGA-3′; IL-12A, 5′-ATGATGGCCCTGTGCCTTAG-3′, 5′-TCCGGTTCTTCAAGGGAGGA-3′; IL-23A, 5′-CTCTGCTCCCTGATAGCCCT-3′, 5′-TGCGAAGGATTTTGAAGCGG-3′. All real-time PCR data analyses were performed using Rotor-Gene 6000 series software. The relative fold changes in expression were calculated using the 2^−ΔΔCt method.

Purified CD5^high and CD5^low DCs at 2 × 10⁵/ml were stimulated in U-bottom 96-well plates for 24 h in complete medium alone or in the presence of polyinosinic-polycytidylic acid [poly(I:C); 10 μg/ml] or R848 (1 μg/ml). Culture supernatants were harvested, and ELISA was performed to detect human IL-6 (3460-1H-20; Mabtech) and TNF-α (3510-1H-20; Mabtech) according to the manufacturer’s instructions. To determine the mRNA levels of cytokines, purified CD5^high and CD5^low DCs were stimulated with poly(I:C) (10 μg/ml; InvivoGen) or R848 (1 μg/ml; InvivoGen) for 6 h, and then the cells were collected for RNA extraction and RT-PCR.

DC chemotaxis experiments were performed in 24-transwell plates with 3-μm pore size polycarbonate filters (Costar; Corning). CD5^high and CD5^low DCs were sorted according to the above protocol. Subsequently, these two DC populations were mixed at a 1:1 ratio and stained with a Brilliant Violet 421–labeled anti-human CD5 Ab (which displays stable fluorescence) at 4°C for 30 min. The mixed DCs were cultured overnight, followed by the evaluation of cell migration using transwell chemotaxis assays. Briefly, the lower chamber was filled with 600 μl of medium supplemented with 20 ng/ml CCL21 (PeproTech), and 10⁵ cells in 100 μl of medium alone were added to the upper chamber. After incubation for 2 h at 37°C, the cells in the lower chamber were harvested using PBS containing 5 mM EDTA. The proportion of CD5^high DCs in the lower chamber was determined by flow cytometry. The proportion of CD5^high DCs before migration was used as the control value.

Human CD4⁺CD45RA⁺ naive T cells were purified from PBMCs using a human naive CD4⁺ T cell isolation kit (Miltenyi Biotec). Sorted CD5^high or CD5^low DCs in different quantities were cocultured with 5 × 10⁴ allogeneic naive T cells labeled with CFSE in 96-well U-bottom plates in 200 μl of complete medium. After 5 d in culture, the percentage of CFSE⁻ T cells was determined by flow cytometry.

Purified DC populations were cocultured with allogeneic naive CD4⁺ T cells labeled with CFSE at a ratio of 1:10 in 96-well U-bottom plates for 5 d. Afterward, the cells were restimulated for 5 h at 37°C with 50 ng/ml PMA (Sigma-Aldrich) and 1 μg/ml ionomycin (Sigma-Aldrich) in the presence of 10 mg/ml brefeldin A (BioLegend) for the final 3 h. Then, the Th cell–specific cytokines IFN-γ, IL-22, IL-4, IL-17, and IL-10 were detected via intracellular staining. Briefly, the cells were incubated with a PerCP-Cy5.5–labeled anti-human CD3 Ab and the yellow fluorescent reactive Live/Dead cell dye (Life Technologies) at 4°C for 30 min. After washing, the cells were fixed with fixation/permeabilization solution (BD Biosciences) on ice for 20 min and were permeabilized in BD Perm/Wash Buffer (BD Biosciences) according to the manufacturer’s protocol. Then, the cells were incubated for 30 min at 4°C with the following anti-human mAbs: IL-22–PE, IL-4–allophycocyanin, IL-10–PE-Cy7, IL-17–allophycocyanin-Cy7, and IFN-γ–Brilliant Violet 421. Detailed information about the Abs used is presented in Supplemental Table I. To determine the mRNA levels of T cell–related cytokines, the cocultured T cells were stimulated with PMA (50 ng/ml) and ionomycin (1 μg/ml) for 3 h; then, the cells were collected for RNA extraction and RT-PCR.

Three biological replicates of CD5^high and CD5^low DCs were purified as described above. PMBCs (3 × 10⁸) for each of the three donors were used to sort the cDC2 subset for the RNA sequencing (RNA-seq) experiments, and 2 × 10⁵ DCs of each subset were recovered after purification for the RNA extraction. RNA extraction and sequencing were performed by BGI Tech (Shenzhen, Guangdong, China). A cDNA library was prepared using 200 ng of RNA and the Illumina TruSeq RNA sample preparation v2 kit. The amplified cDNA library was sequenced using an Illumina HiSeq 2000 sequencer. The accession number of the RNA-seq data is GSE77649 (http://www.ncbi.nlm.nih.gov/geo).

The raw RNA-seq data were aligned to the reference genome (hg19) using TopHat (28). The reference genome sequences used were downloaded from the iGenome ftp site. The reads mapped to each gene were counted by HTSeq (29), and a gene annotation file was provided by GENCODE (release 19) (30); this file was first filtered to retain annotations for protein-coding genes. Low expressed genes were filtered out when they matched anyone of the following two criteria: counts per million <2 in more than three samples, or fragments per kilobase of exon per million fragments mapped reads <1 in more than three samples. The differential expressed genes were mined in the R environment mainly with the edgeR (31) package. Genes displaying a fold change in expression of >1.5 and a false discovery rate of <0.001 were considered to be differentially expressed. Because the six samples of cDC2s were from three independent subjects, donor effects were considered. According to the above criteria, 232 genes were identified as differentially expressed between CD5^high and CD5^low DCs. For gene set enrichment analysis, we used the defined signature gene sets for human CD1c⁺ DCs and CD14⁺ monocytes to analyze the distribution of these signature gene sets in CD5^high and CD5^low DCs (27). We filtered out those genes that were not included in our gene lists. Barcode plots for signature gene set analyses were depicted using the limma package in Bioconductor. Statistical tests of enrichment were performed using the Rotation Gene Set Tests method in limma with 9999 rotations (32). One-sided p values are reported.

RNA-seq data of mouse Esam^high and Esam^low DCs (33) were downloaded from Gene Expression Omnibus under accession number GSE76132. Analysis is started from the raw counts file, and differentially expressed genes are obtained following the same pipeline as our human CD5^high and CD5^low DC analysis. In total, 233 genes were found to be more highly expressed in mouse Esam^high cDC2s, whereas 310 genes are more highly expressed in mouse Esam^low cDC2s. Mouse genes were mapped to human genes using the BioMart database, and only genes with a one-to-one correspondence were kept. In total, 205 (88.0%) Esam^high cDC2 highly expressed genes can be mapped to humans, and 244 (78.7%) Esam^low cDC2 highly expressed genes can be mapped to humans. These mapped genes were used in a barcode plot.

Gene ontology (GO) analysis of the RNA-seq data were performed using DAVID 6.7 (34, 35).

All graphs presented in this article were generated using GraphPad Prism software. Data are shown as the mean ± SEM of independent experiments. Statistically significant differences were determined by paired two-tailed t tests. A p value <0.05 was considered statistically significant.

It has previously been reported that subpopulations of human DCs express CD5 (36, 37). In this study, we investigated CD5 expression by human blood cDC2s after gating for Lin⁻HLA-DR⁺CD123⁻CD11c⁺CD1c⁺ cells. The results showed that a subpopulation of cDC2s expressed significantly higher levels of CD5 than another subpopulation (Fig. 1A). Thus, we referred to these two cell populations as CD5^highCD1c⁺ (CD5^high) and CD5^lowCD1c⁺ (CD5^low) DCs, respectively. The CD5^high DCs constituted ∼25% of all cDC2s (Fig. 1B, n = 15). To confirm that CD5 is actually expressed on the surface of DCs and is not acquired from surrounding cells, we sorted CD5^high and CD5^low DCs and quantified their CD5 mRNA levels via RT-PCR. As shown in Fig. 1C, CD5^high DCs expressed much higher levels of CD5 mRNA (∼10-fold) than did CD5^low DCs. Using purified DCs, we also analyzed the cell morphology of both populations. Both CD5^high and CD5^low cells displayed typical cDC morphology without evident differences (Fig. 1D). Human CD1c⁺ cDC2s from tonsil tissue also showed differential CD5 expression (Fig. 1E), and the frequency of CD5^high cells among tonsil cDC2s was similar to that among PBMCs (Fig. 1F). Next, we analyzed the expression of lineage markers of cDC1s, cDC2s, pDCs, and monocytes on CD5^high and CD5^low DCs (4, 38). The results showed that both CD5^high and CD5^low DCs express the DC-specific marker FLT3 and the cDC2-specific marker CD1c. Additionally, CD5^high and CD5^low DCs did not express specific markers of other lineages such as CD303 (pDCs) and CLEC9A (cDC1s) (Fig. 1G). Furthermore, both CD5^high DCs and CD5^low DCs express much lower CD123 (pDCs) and CD14 (monocytes) (Fig. 1G). In summary, human cDC2s both in blood and in lymphoid organs encompass CD5^high and CD5^low subsets.

Because DCs recognize specific patterns of microbial components, termed pathogen-associated molecular patterns (PAMPs), via pattern recognition receptors (39), we analyzed the expression of TLR2, TLR3, TLR4, TLR7, TLR8, and TLR9 by RT-PCR in CD5^high DCs, CD5^low DCs, pDCs, cDC1s, and monocytes. The results showed that TLR7 and TLR9 are preferentially expressed in pDCs, TLR3 is highly expressed in cDC1s, and TLR2, TLR4, and TLR8 are highly expressed in monocytes (Supplemental Fig. 1A). The expression patterns of TLR2, TLR3, and TLR9 were also confirmed at protein level by FACS (Supplemental Fig. 1B). Thus, the CD5^high and CD5^low DCs have a similar pattern of TLR expression, which is different from the TLR pattern in pDCs, cDC1s, and monocytes. As previously reported (8), cDC1s produced more IFN-β and IL-12A than did CD5^high DCs by poly(I:C) (a TLR3 ligand) stimulation (Supplemental Fig. 1C). Additionally, monocytes are more responsive to R848 (a TLR7 and TLR8 ligand) stimulation than are both CD5^high and CD5^low DCs (Supplemental Fig. 1D).

When comparing CD5^high and CD5^low DCs, we found that CD5^high DCs express a higher level of TLR3 and a lower level of TLR7 and TLR8 (Fig. 2A). Accordingly, CD5^high DCs produced more TNF-α and IL-6 than did CD5^low DCs upon poly(I:C) stimulation (Fig. 2B). Additionally, CD5^high DCs also expressed more IFN-β, IL-12A, and IL-23A mRNAs than did CD5^low DCs upon poly(I:C) stimulation (Fig. 2C). Conversely, CD5^high DCs produced less TNF-α and IL-6 than did CD5^low DCs upon stimulation with R848 (Fig. 2D). Additionally, CD5^high DCs also expressed fewer IFN-β and IL-12A mRNAs than did CD5^low DCs upon R848 stimulation (Fig. 2E). Interestingly, in contrast to their lesser expression of TLR7 and TLR8, CD5^high DCs expressed more IL-23A mRNAs than did CD5^low DCs upon R848 stimulation (Fig. 2E). The mechanisms underlying the discrepancy between TLR7 and TLR8 and IL-23A expression by R848 stimulation need to be further investigated. Overall, these results demonstrated that CD5^high and CD5^low DCs responded differently to PAMP stimulation.

To further characterize CD5^high and CD5^low DCs, we purified these two subpopulations and conducted transcriptome analysis. Pairwise comparison of CD5^high DCs with CD5^low DCs showed 232 genes that were differentially expressed by >1.5-fold (Fig. 3A). Among these, 78 genes were upregulated (Table I, data not shown) and 154 genes were downregulated in CD5^high DCs compared with CD5^low DCs (Table II, data not shown). GO analysis indicated that CD5^low DC highly expressed genes were associated with response to wounding or inflammatory response (Fig. 3B). However, CD5^high DCs highly expressed genes did not have significant enrichment of any GO identifiers (data not shown). Interestingly, we observed that CD5^high DCs expressed more IRF4 but less MAFB than did CD5^low DCs.

IRF4 is essential for the development and function of mouse CD11b⁺ cDC2s (the counterparts of human CD1c⁺ cDC2s) (23–26, 40), and MAFB is a lineage-specific transcription factor that drives monocyte/macrophage development (41–43). Consistent with their elevated MAFB expression, CD5^low DCs also displayed increased levels of monocyte/macrophage-specific molecules such as CD14, CD163 (44), FCN1 (45), S100A8, S100A9 (46), C5AR1 (47), MRC1 (48) and TREM-1 (49) (Table II). We confirmed the expression of S100A9, a calcium-binding protein of the S100 family that plays a pivotal role in the inflammatory response, by flow cytometry analysis (Fig. 3C) (46). We also confirmed the differential expression of IRF4 and MAFB between CD5^high and CD5^low DCs using RT-PCR (Fig. 3D).

Interestingly, mouse CD11b⁺ cDC2s have previously been classified into an Esam^high subgroup that displays a DC signature and an Esam^low subgroup that displays a monocyte signature (15). Subsequently, we analyzed the transcriptomes of CD5^high and CD5^low DCs to determine whether these cells also display DC and monocyte diversification. In this analysis, we took advantage of the defined signature gene sets for human CD1c⁺ cDC2s and CD14⁺ monocytes (27). We found that the genes highly expressed in monocytes were enriched in CD5^low DCs (Fig. 3E, red bars). Additionally, the genes specifically less strongly expressed in monocytes were more strongly expressed in CD5^high DCs (Fig. 3E, blue bars). Furthermore, the expression of CD1c⁺ DC signature genes displayed the opposite trend to monocyte signature genes in CD5^high and CD5^low DCs (Fig. 3F). We also analyzed the expression of defined signature gene sets for human cDC1s and pDCs (27) in CD5^high and CD5^low DCs. We found that the expression of pDC signature genes was higher in CD5^high than CD5^low DCs, but this difference was not statistically significant (p = 0.063) (Supplemental Fig. 2A). The expression of cDC1 signature genes did not show significant difference between CD5^high cDCs and CD5^low cDCs (Supplemental Fig. 2B). This analysis confirmed that both CD5^high and CD5^low DCs are cDC2s. Thus, these data suggested that similar to their counterpart CD11b⁺ cDC2s in mice (15), human blood CD1c⁺ cDC2s can be separated into a CD5^high subpopulation displaying a cDC2 signature and a CD5^low subpopulation displaying a more monocyte signature.

We further analyzed the possible parallel relationship between human CD5^high and CD5^low DCs with mouse Esam^high and Esam^low cDC2s. Esam^high and Esam^low cDC2 highly expressed genes were obtained from RNA-seq data (33). We found that genes highly expressed in Esam^high cDC2s were enriched in CD5^high DCs (p = 0.02) and genes highly expressed in Esam^low cDC2s were also enriched in CD5^low DCs (p = 0.01) (Supplemental Fig. 2C). Reciprocally, genes highly expressed in CD5^low DCs were enriched in Esam^low cDC2s (p = 0.004). However, genes highly expressed in CD5^high DCs were not enriched in Esam^high cDC2s (p = 0.16) (Supplemental Fig. 2D). Overall, these data suggested that human blood CD5^high DCs are more similar to mouse Esam^high cDC2s, whereas human CD5^low DCs are more similar to mouse Esam^low cDC2s. However, whether CD5^high and CD5^low DCs are the counterparts of mouse Esam^high and Esam^low cDC2s needs to be further investigated.

DCs can migrate to draining LNs via afferent lymphatic vessels under the control of the chemokine receptor CCR7. The ability of DCs to migrate to LNs is fundamental for maintaining immunotolerance or initiating an immune response. Mature DCs display very high expression of CCR7 (50, 51), which directs the migration of these cells toward CCL21-expressing lymphatic vessels and draining LNs (17, 18, 52). Freshly isolated CD5^high and CD5^low DCs expressed low levels of CCR7 (Fig. 4A, left panel). Both CD5^high and CD5^low DCs showed upregulated CCR7 expression after culture in vitro (Fig. 4A, middle panel), and CD5^high cells expressed higher levels of CCR7 than did CD5^low DCs. Stimulation with R848 increased CCR7 expression by both DC subpopulations but minimized the difference in CCR7 expression between these subpopulations (Fig. 4A, right panel). Subsequently, we examined the migratory activity of both DC subpopulations using a transwell chemotaxis assay in the presence of CCL21. The results showed that CD5^high DCs migrated toward CCL21 more efficiently than did CD5^low DCs (Fig. 4B). Interestingly, upon stimulation with R848, CD5^high DCs migrated also more efficiently than CD5^low DCs, despite the similar levels of CCR7 expression between these subpopulations (Fig. 4C). This result might reflect the differential expression of other molecules that contribute to cell migration. In summary, the above results suggest that CD5^high DCs are more prone to migrate to draining LNs. To confirm this hypothesis, we analyzed both the migratory (HLA-DR^highCD11c⁺) and resident (HLA-DR^lowCD11c⁺) subpopulations of the CD1c⁺ cDC2s in LNs (12, 27). Approximately 20% (19 ± 3%) and 40% (39 ± 9%) of the CD5^high DCs were observed among resident cDC2s and migratory cDC2s, respectively (Fig. 4D, 4E). Taken together, these results demonstrated that CD5^high cDC2s possess stronger migratory abilities and were overrepresented among migratory cDC2s in LNs compared with blood- or LN-resident cDC2s.

Next, we characterized the capacities of CD5^high and CD5^low DCs to induce naive CD4⁺ T cells proliferation. We first analyzed the expression of MHC class II and costimulatory molecules by the two DC subpopulations. We observed that CD5^high and CD5^low DCs freshly isolated from human blood expressed comparable levels of CD80, CD86, and HLA-DR (Fig. 5A). However, the expression level of LAMP3 (CD208, DC-LAMP), a DC maturation marker (53), was higher in CD5^high DCs than in CD5^low DCs (Table I). Additionally, after culturing in medium, CD5^high DCs expressed higher levels of CD86 and HLA-DR than did CD5^low DCs (Fig. 5B). This spontaneous maturation could reflect DC–DC interactions under the condition of high DC density (54). Moreover, upon stimulation with R848, CD5^high DCs also expressed higher levels of CD86 and HLA-DR than did CD5^low DCs (Fig. 5C). To directly compare the immunostimulatory capacities of CD5^high and CD5^low DCs, we cocultured sorted CD5^high or CD5^low DCs with CFSE-labeled allogeneic CD4⁺ naive T cells at various stimulator/responder ratios. The dilution of CFSE in CD4⁺ T cells was assessed using flow cytometry (Fig. 5D). The results showed that CD5^high DCs induced stronger T cell proliferation at a lower DC/T cell ratio (Fig. 5D, 5E). These data indicated that CD5^high DCs were more potent APCs than were CD5^low DCs and exhibited a robust capacity to induce allogeneic CD4⁺ naive T cell proliferation.

We have demonstrated that CD5^high DCs possessed stronger migratory abilities and induced stronger CD4⁺ naive T cells proliferation. Subsequently, we examined whether CD5^high and CD5^low DCs exhibited different capacities to polarize CD4⁺ T cells. We cocultured CD4⁺ naive T cells with either CD5^high or CD5^low DCs and restimulated the primed T cells with PMA and ionomycin. We detected the expression of the subtype-specific cytokines IFN-γ, IL-10, IL-22, IL-4, and IL-17A via intracellular staining (Fig. 6). The results showed that CD5^low DCs induced the differentiation of IFN-γ–producing T cells (44.5 ± 19.3%) more strongly than did CD5^high DCs (31.8 ± 17.2%). Conversely, CD5^high DCs induced the differentiation of IL-10–producing T cells (3.9 ± 3.9%) more strongly than did CD5^low DCs (1.4 ± 1.5%). Moreover, CD5^high DCs induced more regulatory T cells (Foxp3⁺CD25⁺) than did CD5^low DCs (Supplemental Fig. 3A, 3B). Additionally, CD5^high DCs also induced more IL-22–producing T cells (8.7 ± 3.1% versus 4.7 ± 1.8%), IL-4–producing T cells (8.8 ± 5.8% versus 6.2 ± 4.8%), and IL-17–producing T cells (0.6 ± 0.5% versus 0.4 ± 0.3%) than did CD5^low DCs. We confirmed the differential expression of cytokines at RNA levels by RT-PCR (Supplemental Fig. 3C). We also analyzed the expression of T cell subset–specific transcription factors, including T-bet (Th1), GATA3 (Th2), and RORC (Th17), but we did not observe any significant difference between T cells cocultured with CD5^high and CD5^low DCs (data not shown). The exact reasons underlying the discrepancy between cytokines production and the expression of lineage-specific transcription factors is currently unknown and need to be investigated in the future.

Thus, we showed that the two DC subpopulations polarized CD4⁺ naive T cells into distinct Th cell subtypes. CD5^high DCs induced the polarization of more IL-10–, IL-22–, IL-17A–, and IL-4–producing Th cells but fewer IFN-γ–producing Th cells than did CD5^low DCs.

In the present study, we demonstrated that human cDC2s are heterogeneous. In both blood and lymphoid organs, cDC2s could be separated into two subpopulations based on their CD5 expression. These two DC subpopulations display considerable differences in gene expression and in functions, including cytokine production, migration, T cell activation, and T cell polarization. CD5^high DCs express higher levels of IRF4, which regulates gene expression in a concentration-dependent manner (55, 56). Thus, we reasoned that the expression level of IRF4 might substantially contribute to the phenotypic and functional differences between CD5^high and CD5^low DCs.

IRF4 inhibits cytokine production by competing with IRF5 in interacting with MyD88 (57). Additionally, IRF5 is activated by TLR7/8 signaling but not TLR3 signaling (58). Thus, in addition to the differential expression of TLR3 and TLR7/8 between CD5^high and CD5^low DCs, IRF4 may contribute to the differential responsiveness of CD5^high and CD5^low DCs to poly(I:C) or R848 stimulation. Additionally, IRF4 in DCs inhibits IL-12 production and subsequently suppresses Th1 polarization (59), consistent with the greater induction of Th1 polarization by CD5^low DCs, which express lower levels of IRF4.

IRF4 is essential for the accumulation of migratory DCs in draining LNs, and this process likely reflects the regulation of CCR7 expression and increased DC survival in nonlymphoid tissues (23). We showed that cultured CD5^high DCs expressing elevated levels of CCR7 were more prone to migrate toward CCL21. We also showed that CD5^high DCs were overrepresented among migratory cDC2s than blood- or LN-resident cDC2s. Thus, we suggest that the increased CCR7 expression and migratory capacity of CD5^high DCs might be attributed to their increased IRF4 expression. CD5^high DCs, which exhibit robust migratory abilities and induce IL-10–producing T cell polarization, may play important roles in inducing and maintaining immunotolerance under steady-state conditions (60).

Mouse splenic CD11b⁺ cDC2s can be separated into Notch-dependent Esam^high cDC2s and Notch-independent Esam^low cDC2s (15). Esam^low cDC2s express monocyte-specific genes such as CD14, CX3CR1, and LYZ. Esam^low cDC2s also express higher levels of TLR2 and TLR9 and show increased cytokine production after stimulation. However, Esam^high cDC2s are more potent in priming naive CD4⁺ T cells. In this study, we showed that human blood CD5^low DCs expressed higher levels of CD14, TLR2, TLR9, CX3CR1, LYZ, and CCR2 (Supplemental Fig. 2E, 2F, Table II) and that human CD5^high DCs efficiently induced naive CD4⁺ T cell proliferation (Fig. 5). Our transcriptome analysis suggested the possible similarity between CD5^high and CD5^low DCs with mouse Esam^high and Esam^low cDC2s. However, if CD5^high and CD5^low DCs are the counterparts of mouse Esam^high and Esam^low, cDC2s need to be further investigated.

It was recently reported that human blood cDC2s, but not cDC1s or monocyte-derived DCs, exhibit immune-regulatory properties in the presence of Escherichia coli stimulation (61). In the present study, we showed that CD5^high DCs migrated more efficiently and preferentially induced more IL-10–producing T cells but fewer IFN-γ–producing T cells than did CD5^low DCs. Notably, human blood cDC2s have been used as a cancer therapy in clinical trials (62, 63). Thus, the regulatory properties of the CD5^high subpopulation should be considered.

We thank Juan Yu of the Institute of Biophysics and Xiannian Zhang of Peking University for help in transcriptome analysis.

The authors have no financial conflicts of interest.