Identification of Lineage Relationships and Novel Markers of Blood and Skin Human Dendritic Cells Free

Dendritic cells (DCs) are a family of professional APCs that form an important link between the innate and adaptive immune systems. They are found as specific subsets in tissue and blood, and are of either myeloid or plasmacytoid origin. In their immature form, blood and tissue myeloid DCs bind foreign Ags by an array of C-type lectin receptors (CLRs) expressed on their surface. After exposure to foreign Ags or proinflammatory cytokines, DCs mature and migrate to the draining lymph nodes to present MHC class II–bound foreign Ag to, and activate, T cells. Plasmacytoid DCs (pDCs) are found mainly in the blood and lymph nodes, and function primarily to provide antiviral defense by secretion of very large quantities of IFN-α after migration to areas of foreign Ag exposure or inflammation, although in this setting, they can also present Ag and activate T cells (1).

Myeloid DCs can be further divided into functional subsets based on anatomical distribution and the expression of cell surface markers. Classical blood myeloid DCs express CD11c and CD1c (BDCA1), and a CD141 (BDCA3)-expressing subset equivalent to mouse CD8⁺ DCs has also been defined (2). DC-like blood cells that express CD16 and M-DC8 (3, 4) have been recently classified within the monocyte population (2), although it is clear that there are distinct functional differences (5). In skin, there are at least three DC subsets: two found within the dermis that express either CD1a or CD14, and an epidermal Langerhans cell (LC) expressing CD1a. In mice, there is an additional langerin-expressing dermal DC (DDC) that expresses CD103 (2), but no human counterpart has yet been identified. It is likely that, in time, these DC subsets will be further divided based on the discovery of new novel expression markers.

Although the hemopoietic origin of DCs is clear, the precise relationship of circulating precursors to tissue DCs and the ontogeny of tissue and blood DCs is less well defined in humans. Skin DC subsets have been proposed to originate from both monocyte precursors and committed local DC precursors largely based on the difference in murine skin DCs in wild type mice and those deficient for CSF-1 (6) or its receptor (7), and on human transendothelial migration models (8, 9). Common myeloid precursors have been identified in mice (10, 11), and some studies of human and mouse skin have suggested that CD14⁺ monocytes are the direct precursors of epidermal LCs (12). It has also been suggested that CD16⁺ blood DCs, including those expressing the marker M-DC8 (4), may be immediate precursors of some tissue DCs (8, 13). However, reconstitution of skin DCs after bone marrow transplantation (14) has suggested that LCs are replenished from long-lived, locally proliferating precursors, and that CD1a⁺ DDCs arise from precursors distinct from LCs (15, 16). The relationship between the human CD11c⁺ CD1c⁺ blood myeloid DCs as precursors of tissue DC subpopulations has been unclear, but in murine models, the development of committed DC precursors in blood and tissue from common myeloid precursors and differentiation in lymphoid tissue is now clear (17).

Isolation of ex vivo DC subsets is problematic because they exist in very low numbers (<1% of human skin and blood), and skin DCs are inherently difficult to isolate as immature cells because they are prone to maturation as a result of extraction (18). For these reasons, model skin DCs are extensively used for studies of DC function and viral infection. The most common model, monocyte-derived DCs (MDDCs), can be produced in large numbers of immature cells by culturing CD14⁺ monocytes in IL-4 and GM-CSF (19, 20). More recently, a model LC has been proposed that is derived from the leukemia-derived cell line MUTZ3 (21). Although it is proposed that MDDCs and MUTZ3 LCs most closely resemble CD14⁺ DDCs and LCs, respectively, the relevance of these two model systems remains unclear.

In this study, we aimed to investigate the lineage relationships between ex vivo–derived blood and skin monocytic and DC subsets compared with in vitro–derived model MDDCs and MUTZ3 LCs. We primarily used cells directly isolated without culture, but also used cells from skin and blood that were cultured for 24 h after isolation to allow a more direct comparison with the in vitro–cultured MDDC and MUTZ3-LC. We initially conducted polygenetic analysis on gene expression profiles derived by gene arrays of these cell types. To identify novel markers differentiating between DC subsets, we then used quantitative PCR (QPCR) to measure gene expression profiles of surface proteins. We particularly focused on pattern recognition receptors (PRRs) such as TLRs and CLRs because these are involved in specific detection of pathogen molecules and are expressed in unique combinations by different cell subsets to allow them to best recognize pathogens in specific locations in the body. We also focused on the galectins because they are known to be expressed by leukocyte subpopulations and play an important role in the regulation of the immune response (22, 23).

As previously described (24), skin was separated from s.c. adipose tissue and cut into strips 1.5–2 cm in width. The skin strips were kept in medium with antibiotics at 4°C for 0.5–1 h, and split skin was obtained using a skin graft knife. The split skin was placed in RPMI 1640 (Invitrogen) supplemented with 10% human AB serum (RF10; Sigma-Aldrich) with 4 mg/ml dispase (Worthington) at 4°C overnight. The split skin was washed in PBS and split into dermis and epidermis using fine forceps. Dermal tissue was cut into 1- to 2-mm blocks using scalpels in a scissoring action, and the dermal blocks were placed in 10 ml RF10 containing 4 mg/ml collagenase II (Worthington) and incubated at 37°C with agitation for 1 h. DNase I 75 μg/ml (Roche) was added during the last 30 min of incubation and the tissue repeatedly aspirated through a cutoff Pasteur pipette every 5 min. After incubation, the cells were diluted to 50 ml with PBS at room temperature and passed through a 70-μm mesh and pelleted. The cell pellet was washed once in FACS wash (PBS with 1% FCS and 2 mM EDTA) and labeled with directly conjugated Abs to HLA-DR, CD14, and CD1a. The high HLA-DR–expressing CD14⁺ and CD1a⁺ cells were isolated by FACS as previously described (25). The epidermal sheets were either incubated with 0.3 mg/ml trypsin in RPMI at 4°C for 4–6 h and a single-cell suspension isolated over a Nycodenz gradient, or treated using a similar method to the dermal tissue using collagenase dissociation and DNase labeled with CD1a and HLA-DR before MACS selection using an autoMACS Separator and cell sorting for HLA-DR⁺ CD1a⁺ cells by flow cytometry. Sorted cells were lysed in guanidinium-containing lysis buffer (Qiagen). When analyzing cells after in vitro culture, we performed the earlier procedure after dermal skin sheets had been cultured in media and emigrating cells collected and included in the analysis.

Isolation of blood mononuclear cell populations and DCs was done as previously described (26). Buffy coats were obtained from the Red Cross Blood Transfusion Service (Sydney and Melbourne, Australia). PBMCs were isolated over Ficoll Hypaque (GE Healthcare) gradients, and cell populations were isolated by magnetic bead selection and flow cytometry (Fig. 1). In the standard protocol, cells were labeled with mAb to M-DC8 (“slan”) (4) and goat anti-mouse IgM beads (Miltenyi), and positively selected using MACS columns (Miltenyi). The positively selected cells were labeled with conjugated Abs to HLA-DR and CD16, and sorted for CD16⁺, HLA-DR⁺ large cells by high-speed flow cytometry (FACS Vantage DIVA, FACSAria [BD Bioscience], or MoFlo [Dako Cytomation]). The MACS⁻ cells were labeled with hybridoma supernatant specific for CD14 [3C10] and CD14⁺ cells selected using MACS columns. The MACS⁻ [M-DC8⁻ CD14⁻] fraction was further depleted of lineage markers by labeling with hybridoma supernatant specific for CD3 [OKT3], CD8 [OKT8], CD11b [3G8], and CD19 [FMC63], incubated with MACS GAM-IgG beads. The MACS⁺ CD14⁺ cells were further purified by flow cytometry. The MACS⁻ [lineage-negative] cells were labeled with fluorescent Abs to HLA-DR, CD123, and CD1c or CD11c before final selection of myeloid DCs (CD123⁻, CD1c⁺ CD11c⁺) and pDCs (CD123⁺ CD11c⁻ CD1c⁻) using flow cytometry. Cell purity of skin and blood DC populations after cell sorting was between 95 and 99.5% (average 97%), with <2% contamination with other DC populations. Cultured blood DCs and myeloid cells were obtained by 24-h culture of the DCs in RF10, IL-3 (10 ng/ml; R&D Systems), and GM-CSF (40 ng/ml; R&D System) to maintain cell viability.

MDDCs were differentiated from CD14⁺ monocytes as previously described (18, 27, 28). Human CD34⁺ acute myeloid leukemia MUTZ3 cells (provided by S. Santegoets, VU University Medical Centre, Amsterdam, The Netherlands) were cultured in MEM-α containing ribonucleosides and deoxyribonucleosides (Invitrogen) supplemented with 10% conditioned media from the human renal carcinoma cell line 5637 and 20% FCS (JRH Biosciences) at 10⁵ cells/ml. After 7 d, cells were cultured in MEM-α as described earlier but additionally supplemented with 100 ng/ml GM-CSF, 2.5 ng/ml TNF-α, and 5 ng/ml TGF-β1 (R&D Systems) at 2.5 × 10⁵ cells/ml. Cells were cultured for 10 d to allow differentiation into MUTZ3-derived LCs (MUTZ3 LC). The culture media were replaced with fresh cytokine-supplemented media on days 3 and 7. The phenotype of the differentiated MUTZ3 LCs was assessed by flow cytometry. Immature MUTZ3 LCs were defined as cells that stained positively for langerin, CD4, and CD1a, whereas staining negatively for DC-specific intercellular adhesion molecule-3-grabbing nonintegrin (SIGN), mannose receptor (MR), and CD83.

Normal foreskin tissues were obtained from children undergoing circumcision, and mechanically separated into inner and outer parts. The inner foreskin tissues were snap frozen in OCT, cut into 5-μM sections, placed on slides, and kept at −80°C until used for immunofluorescent staining. Tissues were fixed for 10 min with ice-cold methanol/acetone (1:1). Foreskin tissues were blocked with 10% normal goat serum (Sigma-Aldrich) for 30 min at room temperature and then incubated for 45 min at 37°C with rabbit anti-human DC-SIGN polyclonal Ab (1:50; Abcam) and mouse anti-human CD14 mAb (1:20; BioLegend). After the incubation with the primary Abs, Alexa Fluor 546–conjugated goat anti-rabbit (1:400; Molecular Probes) and Alexa Fluor 647–conjugated goat anti-mouse (1:200; Molecular Probes) Abs were added as secondary Abs followed by incubation for 45 min at 37°C. Tissue was incubated for 45 min at 37°C with FITC-conjugated mouse anti-human CD1a Ab (1:10; BioLegend). All washes between steps were carried out in PBS. All Abs were diluted in Protein Block Serum-Free solution (DAKO). After staining, ProLong Gold Antifade reagent with DAPI (Invitrogen) was added to the stained tissues and coverslips were mounted onto tissue-loaded slides. Slides were visualized through a ×40 1.35 NA oil-immersion lens with an inverted Olympus IX-70 microscope (DeltaVision Image Restoration Microscope; Applied Precision/Olympus) and a Photometrics CoolSnap QE camera.

Total RNA was extracted from purified cell populations from individual donors and processed for hybridization to 1 of 55 cDNA gene arrays (Human ResGen 8k; Australian Genome Research Facility) using a common MDDC reference, or 24 bead arrays (sentrix human 6 v2 expression chips; Illumina, San Diego, CA). The RNA extraction, labeling, hybridization, data processing, and analysis procedures are described previously for the cDNA gene array (18) and Illumina arrays (27). Clustered data were further processed in PARTEK Genomics Suite (Partek) to exclude genes not showing detectable expression in >80% of arrays and to remove batch effects. Microarray data are available through the Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE32648 and (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE32400). Genes differentially expressed in at least one group were identified using ANOVA corrected for multiple testing (step-up false discovery rate, p > 0.05), and the data were clustered using City-block (arrays) and average Euclidean (genes) algorithms or by principle components analysis.

Differential expression analysis was initially conducted in Bead Studio using the mean of each cell group as the reference. Gene lists were then generated by filtering the data set for genes absent in the reference group (detection p > 0.05) and genes present in the DC subset group of interest (detection p < 0.01). The SD of the average signal between biological replicates was then calculated. The SD was then divided by the mean average signal to generate a coefficient of variation. Any genes that did not meet the coefficient of variation cutoff were then removed (1.4 for cell sets with 2 biological replicates, 1.7 for 3 biological replicates, and 2.0 for 4 replicates).

QPCR was performed on cDNA derived from the same samples used for microarray analysis, as well as for additional samples using GAPDH as an internal reference for normalization using the methods previously described (27, 28). The primer sequences are listed in Supplemental Table I.

Murine mAbs TLR1-PE (GD2.F4), TLR2-FITC (TL2.1), TLR3-PE (TLR3.7), and TLR4-PE (HTA125) were purchased from eBioscience. Unconjugated TLR5 mAb (19D759.2), biotinylated TLR6 mAb (86B1153.2), unconjugated rabbit polyclonal TLR7, TLR8-FITC (44C143), and TLR9-FITC (26C593.2) were purchased from Imgenex. Goat anti-mouse IgG-PE was from Molecular Probes, streptavidin-PE from BD Pharmingen, and goat anti-rabbit Ig-FITC from Sigma-Aldrich. Cells were stained for surface expression of TLR1, TLR2, TLR4, TLR5, and TLR6 or else fixed and permeabilized with Cytofix/Cytoperm (BD Biosciences), then stained for intracellular expression of TLR3, TLR7, TLR8, and TLR9. TLR5, TLR6, and TLR7 Abs were detected with goat anti-mouse IgG-PE, streptavidin-PE, and goat anti-rabbit Ig-FITC, respectively.

To investigate the lineage relationships between human DC and monocytic cell populations, we isolated DCs and monocytes from human blood and skin using magnetic bead and flow cytometry–based cell sorting (Fig. 1). We obtained representatives of the cells recognized in current classifications (2) and divided the blood cell populations into four groups: 1) CD14⁺ monocytes, 2) CD16/M-DC8⁺ monocytes/DCs, 3) CD11c⁺ blood myeloid or classical DCs, and 4) CD123⁺ blood pDCs. Skin DCs were divided into three groups: 1) CD1a-expressing DDCs, 2) CD14-expressing DDCs (24, 29), and 3) CD1a-expressing epidermal LCs. We also generated the model MDDCs and MUTZ3 LCs in vitro to compare with the ex vivo–derived skin DCs.

We first compared different ex vivo DC and monocytic populations using hierarchical cluster analysis on gene expression profiles detected using cDNA microarrays. Hierarchical clustering uses unique subsets of genes expressed in each sample to determine the degree of similarity between the cell types, providing a representation of the relationship between the samples (Fig. 2A). As expected, pDCs clustered away from the myeloid populations as a distinct unrelated group. The remaining myeloid populations separated into two broad clusters. The first contained all blood cell populations: CD11c⁺ myeloid DCs and CD16⁺ and CD14⁺ monocytes. The second broad cluster contained mixed populations of skin DCs, comprising the CD14⁺ and CD1a⁺ DDCs, as well as epidermal LCs.

Principle component analysis of the expression profiles (Fig. 2B) confirmed the strong separation of blood and skin cells into two broad groups with contained subgroups. In addition, within the skin DC populations, the CD14⁺ DDCs and LCs clustered separately, with the CD1a⁺ DDC cluster overlapping the two populations (not seen in the hierarchical clustering). In addition, in the principle components analysis, cultured cell populations fell into two distinct groups. Cultured CD14⁺ DDCs clustered closely to the skin DC populations, whereas the cultured pDCs and blood myeloid DCs clustered with the blood cell populations (individual cultured cell types are shown in Supplemental Fig. 1).

To further discriminate between the three skin DC subsets and the blood myeloid DC and monocytic cell populations, the clustering analysis was repeated using data generated from Illumina bead arrays that contain complete coverage of the genome. To allow for a comparison of ex vivo– and in vitro–derived DCs, additional populations were profiled, including MDDCs (a model for CD14⁺ DDCs) and MUTZ3 LCs (a model for epidermal LCs). In agreement with the cDNA arrays, the three ex vivo myeloid blood cell populations clustered together, forming a distinct cluster separate from the skin DCs. The three ex vivo–derived skin DC populations still clustered closely, with the CD1a-expressing epidermal LCs and DDCs clustering together and separate to the CD14⁺ DDC population. Consistent with their use as model skin DCs, the in vitro–derived MDDCs and MUTZ3 LCs both clustered with the skin DC cluster. Immature MDDCs clustered most closely to the CD14⁺ DDC population. However, the MUTZ3 LCs formed the most discrete cluster, most similar to mature MDDCs, and did not cluster close to the LC populations.

To explore the lineage relationship of the skin DCs suggested by the clustering analysis, we looked at DDCs before and after culture of intact dermis to determine whether there was any evidence of phenotypic shifts between CD14⁺ and CD1a⁺ DDCs in situ. Before in vitro culture, the high HLA-DR–expressing cells extracted by collagenase treatment included a similar number of CD14⁺ DDCs compared with CD1a⁺ DDCs, but postculture, the CD1a⁺ DDCs were the predominant population in the emigrant and collagenase isolated cells (Supplemental Fig. 2A). To further define the distribution of CD14⁺ and CD1a⁺ DDCs, we cultured dermal sheets before and after collagenase digestion and compared populations as shown in Fig. 3A. The comparison populations were: 1) cells isolated by collagenase digestion at day 0, 2) cells migrating from day 0 collagenase-treated dermis during overnight culture, 3) cells isolated from dermis by collagenase after overnight culture, and 4) cells migrating from intact dermal sheets during overnight culture. Total isolated at day 0 cells = (population 1) + (population 2) and at day 1 = (population 3) + (population 4) (Fig. 3D). Phenotypic analysis of these populations is shown in Fig. 3A. The recovery of CD1a⁺ DDCs was greatest during migration from intact dermal sheets, and this population also had the highest frequency of CD1a⁺ cells (Fig. 3D). The CD1a⁺ cells uniformly expressed HLA-DR in directly isolated cells and in migrating cells, making it unlikely that the increase in CD1a-expressing cells derive from an HLA-DR⁻ population (Supplemental Fig. 2B). High HLA-DR–expressing CD14⁺ cells were highest in cells from dermis collagenase treated at day 0 and in the migrating cells from intact dermis. HLA-DR high, CD14-expressing cells were infrequent among emigrants from collagenase-treated dermis or isolated by collagenase treatment at day 1 of culture. Taken together, these data suggest that there may be phenotypic conversion or maturation of CD14⁺ DDCs during migration. In intact skin, cells coexpressing CD14 and CD1a were observed (Fig. 3B, 3C). We next sorted CD1a⁺, CD14⁺, and dual-expressing cells and cultured them overnight as purified populations (Fig. 3B). Conversion of CD14⁺ to CD1a⁺ DDCs was not observed; however, cells coexpressing CD14 and CD1a did show increased expression of CD1a and reduced CD14. Taken together, the close clustering by gene expression and the apparent development of increased numbers of CD1a⁺ DDCs within cultured explants suggests close lineage relationships between DDC subpopulations and CD14-expressing cells as precursors for the CD1a-expressing cells within the dermal microenvironment in situ.

DC and monocytic cell populations express an array of surface markers that are often unique to specific cell subsets. We identified differential expression of these markers from the microarrays, and confirmed and extended these observations by QPCR using a larger sample size. We focused on PRRs, which are involved in specific detection of pathogen molecules and subsequent signaling. In addition to TLRs and the cytoplasmic RNA binding receptors (RIGI, MDA5, and PKR), we particularly focused on the CLR domain family members (CLECs) that recognize pathogens and vary in their expression on different DC subsets. We also looked at other CLRs (selectins and collectins) and also the galectin family of surface proteins (lectin, galactoside-binding soluble [LGALS]), because these are known to be expressed by various cells of the immune system and participated in immune regulation and homeostasis of various leukocyte subpopulations (22, 23). Finally, we investigated the expression of CD1a and genes encoding costimulatory molecules indicative of DC maturation (CD40, CD80, CD83, and CD86).

The average expression for each cell population is summarized in Table I and illustrated as a heat map (Fig. 4A). Changes in the expression of TLR genes in blood-derived cells were confirmed at the protein level by flow cytometry (TLR1-8; Fig. 4B).

CD1a was exclusively expressed in MDDCs, CD1a⁺ DDCs, and LCs, consistent with the fact that these are the only cell subsets that express this surface marker. LGALS8 was expressed by all cell types, but other galectins were much more cell-type specific. LGALS2 expression was specific to skin cells. High LGALS1 expression was observed only in DDCs. DDCs were the only cells that did not express LGALS9. LGALS7 was expressed only in LCs, which also expressed much lower levels of LGALS4, especially when compared with DDCs. Expression of LGALS12 (GRIP1) was restricted to CD14⁺ monocytes and LGALS10 (CLC) expression was only in pDCs and MDDCs.

Similarly, many of the CLRs were discriminatory for particular cell subsets. First, known patterns of expression were verified such as exclusive expression of CLEC4K (langerin) by LCs (Fig. 4C) and CLEC4C (BDCA2) by pDCs (30).

Recently, langerin-expressing DDCs have been identified in mice (31), but it is unclear whether human dermis contains these cells. We therefore examined the QPCR data more closely to determine whether any other cell populations expressed low levels of langerin (Fig. 4C). The CD1a⁺ DDCs expressed the next highest langerin levels (though very low and below the cutoff for inclusion in Table I). However, no langerin expression was detected in CD14⁺ DDCs. Even lower langerin expression was detected in MDDCs and in five of the eight CD11c⁺ blood myeloid DC donors.

Novel patterns of CLR expression were also defined. CLEC8A (OLR1) was observed only in skin DCs. As expected, CLEC13D (mannose receptor I) and the highly related CLEC13DL were expressed by MDDCs and DDC subsets, as was CLEC3B (TNA), whereas CLEC13E (mannose receptor II) was expressed by MDDCs, CD11c⁺ blood myeloid DCs, LCs, and at very low levels by two of five of the CD1a⁺ DDC samples. CLEC4E (MINCLE) expression was observed only in DDCs, and these cells also expressed higher levels of CLEC4M (LSIGN) than other cell populations. CLEC13B (DEC205) was not expressed by CD14⁺ DDCs and at only very low levels by CD1a⁺ DDCs, which were also the only cells found not to express CLEC12A (DCAL2). CLEC4G (LSECtin), CLEC4L (DCSIGN), and CLEC14A (EGFR-5) were all uniquely expressed by MDDCs and CD14⁺ DDCs. CLEC2B was exclusively not expressed by MDDCs. CLEC5A (MDL1) expression was specific for CD1a⁺ DDCs and LCs. CLEC6A and CLEC7A (Dectin 2 and 1) were not expressed by pDCs, and CLEC1A was not expressed by pDCs and LCs. SELL (CD62L) was not expressed by any skin DC populations or MDDCs, and SELP (CD62) showed variable expressed in most cells but was not expressed by any CD1a⁺ DDC samples. Most collectins were not expressed by any cell populations; however, COLEC12 was expressed very highly by MDDCs and DDCs, and COLEC7 was not expressed by pDCs.

Concerning other PRRs, PKR, RIGI, and MDA5 were consistently expressed across all populations, as were TLR1 and TLR6. TLR5 expression was very low in MDDCs and myeloid blood cells, and absent in all other cells. As expected, TLR7 and TLR9 were expressed much more highly in pDCs, which were the only cells that did not express TLR2 and TLR3. TLR4 and TLR8 were not expressed in DDCs, LCs, and pDCs (one of the CD1a⁺ and CD14⁺ DDC donors expressed TLR4 and TLR8, respectively). Gene and protein expression of TLRs in blood-derived cell subsets was generally consistent, although TLR7 was detected at the gene but not protein level in CD14⁺ monocytes. It is of interest that TLR3 is expressed much more highly on the model MDDCs and TLR4 on CD14⁺ monocytes than the other cell populations.

Consistent with the literature, the ex vivo–derived skin DCs all showed evidence of partial maturation as evidenced by increased expression of CD40, CD83, and CD86 (32). These results are summarized at the top of Table II.

We next examined the Illumina gene expression data to identify genes whose expression was unique to specific DC populations. Those with the most restricted expression were tested by QPCR (Fig. 5). We identified specificity of expression for the following genes and cell types: CCL18 in MDDCs, SLC11A1 in all monocytes, SMARCD3 in CD14⁺ monocytes, LYPD2 and PTPN3 in CD16⁺ monocytes, SLA2 in CD1a⁺ DDCs, PRSS2 and PEEP1 in CD14⁺ DDCs, and EFNB3 in LCs. In addition, we identified lack of expression for the following genes and cell types: TNI2 not in MDDCs, INF2 not in CD16⁺ monocytes, ACOT7 not in myeloid blood cells, STK11 not in blood myeloid DCs, PLXNC1 not in CD1a⁺ DDCs, and PKN not in LCs. These results are summarized at the bottom of Table II.

Functional clustering of genetic markers represents a way of studying lineage and differentiation of highly purified cell populations (33). Microarray analysis remains one of the most powerful tools for the identification of such lineage relationships especially among blood cells and malignancies (34–38). In this study, we have used hierarchical clustering and principle component analysis to further understand the origins and relationships between isolated blood and skin DC and monocytic cell populations (Fig. 1). We have gone on to identify novel markers to help further distinguish between different cell populations using QPCR.

Hierarchical clustering and principle component analysis (Fig. 2A, 2B) showed that pDCs formed the expected discrete cluster because of well-established early lineage differences from other myeloid cells. The myeloid skin and blood populations also clustered separately, indicating no clear similarity of gene expression profiles between these two general cell populations. These data did not provide direct support for the model of CD11c⁺ blood myeloid DCs (or a subset of them) as direct precursors to skin DCs in the resting state. However, treatment of blood CD14⁺ monocytes in vitro with cytokines differentiates them into MDDCs that cluster with skin DCs demonstrating the plasticity of these lineages. Principle component analysis and Illumina expression array data (Fig. 2B, 2C) revealed individual subsets within these two main populations that clustered apart from each other. Within the skin, the CD1a⁺ DDCs and LCs clustered most closely, consistent with the common expression of CD1a on their surface, with the CD14⁺ DDCs forming a separate cluster. When we investigated the effects of skin explant culture on isolated DDCs in situ, we saw a phenotypic shift from CD14-expressing cells to CD1a-expressing cells, with some cells expressing both markers (Fig. 3C, 3D). In addition, we found that after overnight culture, the proportion of CD1a⁺ cells isolated increased and CD14⁺ cells decreased. These observations of close clustering of skin DCs and a phenotypic shift after culture suggest that the dermal skin populations may be related and interconvert from CD14⁺ to CD1a⁺ DDCs without direct origin from separate blood monocytic precursors. However, we did not observe a conversion of CD14⁺ to CD1a⁺ cells after purification, indicating that a factor within the microenvironment of the skin is necessary to drive the conversion. Our findings in this study of expression profiles of segregated populations based on anatomical location (blood versus skin) are consistent with a recent murine DC microarray study that looked at lymphoid resident DCs (39).

Recently, langerin-expressing CD103⁺ DDCs have been discovered in mice (31). Langerin-expressing DDCs have also been reported in human skin; however, these have been considered as epidermal LCs in transit through the dermis (40), and there remains speculation as to the existence of an equivalent population in humans. Although we found that LCs were the only cell population that expressed high langerin levels, we did find that CD1a⁺ DDCs (but not CD14⁺ DDCs) also expressed langerin, albeit at very low levels (Fig. 4C). We cannot completely exclude the possibility of minor contamination between CD1a-expressing LCs and DDCs. In addition, the principle components analysis showed an overlap between the CD1a⁺ DDC and LC populations, which also clustered closely together in the Illumina expression analysis. This suggests that a human langerin-expressing DDC subset may exist within the CD1a⁺ DDC population but represent only a trace population in resting skin.

The differentiation of CD14⁺ monocytes into MDDCs using IL-4 and GM-CSF (19, 20) has provided a method for generating large numbers of DCs for immunotherapy and a model for the study of DC biology in vitro. Similarly, LC-like cells can be generated in vitro either from CD34⁺ bone marrow or blood precursors cultured in GM-CSF and TNF-α (41), or from CD14⁺ monocytes using TGF-β, GM-CSF, and IL-4 (42). Although human monocytes have been shown to differentiate into both macrophages and DCs after transendothelial migration (13), and a TGF-β–dependent pathway from CD14⁺ monocytes to LCs may exist in vivo (12), it is unclear what the in vivo equivalents of these model DCs are. MDDCs have been proposed to most closely resemble CD14⁺ DDCs partly because both express the CLR DC-SIGN (25).

LC-like cells derived from CD34⁺ or CD14⁺ precursors are assumed to mimic LCs because of their high langerin expression, but unlike LCs, they also express the CLRs DC-SIGN and MR, similar to MDDCs. In addition to expressing high levels of langerin, MUTZ3 LCs lack the expression of DC-SIGN and MR (21), and for this reason, we chose to investigate this model LC. In support of their use as model skin DCs, both the MDDC and MUTZ3 LC cell populations did reside within the general skin DC cluster (Fig. 2C). Consistent with their proposed similarity to CD14⁺ DDCs, the immature MDDCs clustered most closely to this ex vivo cell type. However, the MUTZ3 LCs cells did not cluster closely to LCs and rather formed the most discrete cluster within the skin cells alongside mature MDDCs, indicating that caution should be applied to findings derived from this model system. However, they would remain a useful model for studies involving surface receptors.

In support of MDDCs as model DDCs, we found common patterns of exclusive CLR expression, both uniquely expressing CLEC13D (mannose receptor), CLEC13DL, CLEC3B (TNA), and COLEC12. Furthermore, in support of a closer similarity to CD14⁺ DDCs than CD1a⁺ DDCs, we confirmed the finding that CLEC4L (DCSIGN) expression was restricted to CD14⁺ DDCs and MDDCs, and we also show that the related CLEC4G (LSECtin) and CLEC14A were also uniquely expressed by these two cell types. Like DC-SIGN, LSECtin is involved in pathogen binding and recognizes endogenous activated T cells (43), and CLEC14A has recently been identified as an endothelial marker involved in cell migration (44). No CLRs were uniquely expressed or nonexpressed by MDDCs and LCs or CD1a⁺ DDCs.

In this study, we have identified additional specific markers for the different DC/monocyte subpopulations (Table II), which may serve as a basis to more easily distinguish between subpopulations: 1) CD14⁺ and CD16⁺ blood monocytes specifically express SLC11A1, SLC5A10, and SMARCD3. SLC11A1 and SLC5A10 are both solute carrier proteins. SLC5A10 has been shown to be exclusively expressed in myeloid lineage cells and plays a role in macrophage differentiation where the recruitment of SWI/SNF complex is required for its transcriptional activation (45). SMARCD3 is a core component of the nuclear SWI/SNF complex linking the cell-type–specific function of the proteins encoded by these two genes. This complex is also required for HIV TAT-driven transcriptional activation (46). 2) CD14⁺ monocytes uniquely express LGALS12, previously been shown to be involved in regulating the cell cycle (47). 3) CD16⁺ monocytes uniquely express LYPD2 and PTPN. No studies have been published on LYPD2, but PTPN3 is a member of the protein tyrosine phosphatase family that regulates a variety of cellular processes including cell growth and differentiation. INF2 was exclusively not expressed by CD16⁺ monocyte. INF2 is involved in immunologic synapse formation (48), and its nonexpression is consistent with limited ability of CD16⁺ monocytes to form immunological synapses compared with DCs or other cells that form such synapses. 4) CD11c⁺ blood myeloid DCs did not exclusively express any genes. However, they alone did not express the serine/threonine kinase STK11. 5) pDCs showed the most marked differences in their surface marker expression compared with other ex vivo cells as expected. They exclusively expressed very high levels of CLEC4C (BDCA2), LGALS10 (CLC), TLR7, and TLR9. In addition, they were the only cells not to express COLEC7, TLR2, TLR3, as well as CLEC6A (dectin 2) CLEC7A (dectin 1), which play a role in innate recognition of fungi (49).

Skin cell populations uniquely express the low-density lipoprotein scavenger receptor CLEC8A (OLR1). Selectin L was not expressed by skin DCs, consistent with an essential role in blood leukocyte emigration through vascular endothelium (50). Presence or absence of some markers were specific for skin DC subpopulations. 1) Epidermal LCs uniquely express very high levels of LGALS7 and langerin, as well as EFNB3, as previously shown by others (51). This is consistent with known expression of LGALS7 by epithelial cells (52). LGALS2 was expressed at much lower levels than other cells. We found that our uncultured LCs did not express TLR4 as previously reported by others (53), and expression of TLR2 was detected in only two of the four LC donors. This is consistent with previous work (54), although a recent publication shows TLR2 and TLR4 expression on LCs after culture (55). LCs did not express TLR5 or TLR9 and uniquely lacked PKN, which regulates NF-κB signaling (integral to TLR-mediated signal transduction) via phosphorylation of TRAF1 (56). 2) DDCs uniquely express LGALS1, which plays a role in leukocyte migration, modulates proinflammatory cytokine release (57), and mediates DC anergy (58), and CLEC4E (MINCLE), which plays a role in recognizing pathogenic fungus (59). LGALS9, which plays a pivotal role in T cell immunity (60), was uniquely not expressed by DDCs. 3) CD1a⁺ DDCs uniquely express SLA2, which has recently been shown to regulate DC maturation (61). In addition, they expressed much higher levels of CLEC4M (LSIGN) shown to be associated with HIV transmission (62). CD1a⁺ DDCs were the only cell population that consistently did not express the monocytic cell marker CELC12A (DCAL2). 4) CD14⁺ DDCs exclusively expressed REEP1, which is involved in targeting receptors to lipid rafts (63). Conversely, PLXNC1 was exclusively nonexpressed. In murine DCs, PLXNC1 signaling inhibits integrins resulting in decreased cell adhesion and migration, and DCs in mice lacking PLXNC1 showed a reduced capacity to stimulate T cells (64).

The ontogeny of DCs, and their migration to and population of tissue have important implications for dissemination of infectious organisms. For example, HIV can readily infect or be carried by blood DCs or CD16⁺ monocytes (26), which marginalize (5) and emigrate under inflammatory conditions to generate tissue DCs. We have previously shown that different CLRs may bind HIV in tissue (25), and recent work has suggested that within the skin, langerin-mediated binding by HIV may have differing effects depending on the state of the cell (65). In this study, we have now extended our understanding of DC–HIV interactions by providing profiles for expression of surface markers on DC subsets in blood and skin that may be important in HIV entry and virus carriage by these DC subsets. Such data are also relevant to other epitheliotropic viruses such as HSV. Understanding these processes can only be achieved by a study of primary cells and use of uniquely expressed and nonexpressed markers as we have identified in this study.

In conclusion, we show in this study that skin and blood DC/monocyte populations form two distinct polygenetic cell populations. This and the finding that CD14⁺ DDCs convert to CD1a⁺ DDCs in situ does not support the hypothesis that CD11c⁺ blood myeloid DCs represent a direct skin DC precursor, but rather indicates that CD14⁺ DDCs may be precursors to CD1a⁺ DDCs. In line with their common expression of CD1a, LCs and CD1a⁺ DDCs had a greater degree of similarity in expression profiles compared with the CD1a-nonexpressing CD14⁺ DDCs. Our findings support the proposal that MDDCs most closely resemble CD14⁺ DDCs. In addition to DCSIGN, we have identified three CLRs that these two cells populations uniquely coexpress. MDDCs probably represent the closest model of DDC, but MUTZ3 LCs do not cluster with LCs and, therefore, may be limited as a LC model. Furthermore, we have identified uniquely expressed and nonexpressed novel genes that discriminate between DC subsets. These findings now provide a broader range of unique markers for fully differentiated human skin and blood DCs, which may help in defining human skin precursors during injury inflammation. They also indicate multiple new directions for research to define further biological differences between DC subsets.

Flow cytometry was performed in the Alfred Monash Research Education Precinct core flow cytometry unit (Melbourne, Victoria, Australia) and the Westmead Research hub Flow Cytometry Centre, supported by Westmead Millennium Institute, Children’s Medical Research Institute, Kid’s Research Institute, National Health and Medical Research Council and Cancer Institute New South Wales. Dr. John Anstee (Mercy Hospital, East Melbourne, Victoria, Australia) and Dr. Norman Oldbourne (Dalcross Adventist Hospital, Killara, New South Wales, Australia) provided discarded skin samples from apronectomy operations.

The authors have no financial conflicts of interest.