Characterization and Transcriptomic Analysis of Porcine Blood Conventional and Plasmacytoid Dendritic Cells Reveals Striking Species-Specific Differences Free

The mononuclear phagocyte system (MPS) is a complex network of innate immune cells specialized in sensing the environment through the expression of a wide range of pattern recognition receptors (PRRs) that recognize conserved microbial-associated molecular patterns. Based on ontogeny, the MPS can be divided into monocytes and cells derived thereof, tissue macrophages and bona fide dendritic cells (DCs) (1). Mouse studies have shown that monocytes and DCs originate from hematopoietic stem cells in the bone marrow that give rise to a monocyte and DC precursor, which further differentiates into monocytes and a common DC precursor, giving rise to the bona fide DC subsets (2–4).

DCs are professional APCs that provide an important link between the innate and adaptive immune response, and shape the immune response depending on the innate signals received and the exposed tissue microenvironment. Three functionally and phenotypically distinct DC subsets have been extensively studied in mice and humans. These include plasmacytoid DCs (pDCs) and conventional DCs (cDCs), the latter being further divided into a cDC1 and a cDC2 subset (3). Although pDCs are particularly important for anti-viral responses through production of high amounts of type I IFNs (especially IFN-α), cDCs are most efficient at presenting Ags and activating naive T cells. In mice, cDC1 are specialized in Ag cross-presentation to CD8 T cells, and are particularly efficient in promoting Th1 responses (3, 5). Functionally, cDC2 have been shown to be efficient at promoting Th17 as well as Th2 immune responses (3, 5). The functions observed for each of these cDC subsets are not mutually exclusive as the cDC network is plastic with overlapping properties. Nevertheless, with respect to innate immune responses, it is also important to note that DC subsets differ in expression of TLRs. Although some patterns of expression are common between mice and humans, such as TLR7 and TLR9 overexpression by pDCs or TLR3 expression mainly by the cDC1 subset, some differences are observed. For example, TLR9 is restricted to pDCs in humans but is also expressed by cDCs in mice (6).

Each subset has first been characterized by their expression of specific markers, some of which are conserved, whereas others vary between species. For example, key markers for cDC1 are CD8α and CD103 in mice, but CD141 in humans. In both species, the expression of the chemokine receptor XCR1 and the C-type lectin receptor CLEC9A is cDC1 specific. Key markers for cDC2 are CD11b⁺ in mice and CD1c⁺ in humans. In contrast to cDC1, cDC2 expresses high levels of CD172a in both species. Important differentiation markers for human pDC are CD303, CD304 and CD123, whereas for mouse B220, siglec H and PDCA-1 are often used (3). As pDC differentiation depends on the expression of the transcription factor E2-2 (TCF4), this gene may be used as a universal pDC marker (7). In conclusion, although commonalities are found between the DC subsets of mice, humans and veterinary animals, it is evident that each species has its own peculiarities in terms of marker expression and subset function (8). Thus, for each species of interest, the distinctive phenotype and functional characteristics of each subset need to be identified. The identification of common gene expression has been used to determine the porcine counterparts of DC subsets and monocytic cell subsets (9–11). Nevertheless, the markers employed remained limited, increasing the risk of cell-type cross-contamination, which would make conclusions on species peculiarities difficult. Therefore, to address this question, unbiased approaches that can cover the whole transcriptome such as next-generation sequencing are advantageous.

In the current study, we employed multicolor flow cytometry (FCM) and identified a panel of 11 mAbs that bind to differentially expressed surface molecules on porcine cDC1, cDC2, pDC, and monocytes, enabling a highly confident and precise definition of subsets in this species. The identity and purity of the subsets was confirmed by transcriptional profiling using RNA sequencing of the sorted subsets. To our surprise, particularly in the porcine pDC subset, striking differences in gene expression and functional responses to TLR ligands when compared with mice and humans were found.

Blood was collected from 3 to 12 mo-old Large White pigs. The animals were kept under specific pathogen-free conditions and blood sampling was approved by the Animal Welfare Committee of the Canton of Bern, Switzerland (animal licenses BE26/11 and BE88/14). PBMCs were isolated using a Ficoll-Paque density gradient (1.077 g/l; Amersham Pharmacia Biotech). Monocyte isolation was performed as previously described (12). Briefly, PBMCs were incubated with anti-human CD14 magnetic beads (Miltenyi Biotec) and CD14 positive cells were sorted on an LS column using the MACS system (Miltenyi Biotec). CD14⁺ monocytes were then differentiated into monocyte-derived macrophages (MDMs) as previously described (13). CD14⁺ monocytes were seeded in 12-well plates with 10⁶ cells/well in 2 ml DMEM with Glutamax (Life Technologies) and 10% heat-inactivated porcine serum (Sigma Chemicals). Cells were cultivated for 3 d at 39°C with 5% CO₂ before being harvested.

The different DC subsets were identified by FCM using a four-step, five-color staining protocol. Unless specified, all Abs used were directed against the porcine molecule. Between each staining step, the cells were washed with Cell Wash (BD Biosciences). PBMCs were first incubated with anti-CD172a (hybridoma clone 77-22-15A, kindly given by Dr. Armin Saamüller, Veterinary University of Vienna, Austria) and anti-SynCAM (TSLC1/CADM1) (cross-reactive anti-mouse Syn-CAM mAb, clone 3E1; MBL) Abs. Cells were then incubated with the secondary Abs anti-mouse IgG2b Alexa Fluor 647 (Molecular Probes) and anti-chicken IgY biotin (Jackson ImmunoResearch). A third step of Ig blocking was then performed using ChromPure mouse IgG (Jackson ImmunoResearch). Cells were finally incubated with the conjugated Abs anti-CD14-FITC (clone MIL2; AbD Serotec) and anti-CD4-PerCP-Cy5.5 (clone 74-12-4; BD Pharmingen), and V500-coupled streptavidin (BD Horizon). The cell-surface marker expression of the different subsets was further characterized using anti-wCD11R1 (clone MIL4; Serotec), anti-CD1 (hybridoma, clone 76-7-4; kindly provided by Dr. Armin Saalmüller, Veterinary University of Vienna, Austria), anti-CD115 (CSF-1R, hybridoma clone ROS8G11-1; kindly provided by Prof. David Hume, Roslin Institute, University of Edinburgh, U.K.), anti-CD163 (hybridoma clone 2A10/11; kindly given by Dr. Javier Dominguez, Instituto Nacional de Investigacion y Tecnologia Agraria y Alimentaria, Madrid, Spain), anti-CD205 (hybridoma clone ZH9F7; kindly provided by Dr. Jesus Hernandez, Centro de Investigación en Alimentación y Desarrollo, Hermosillo, Mexico), anti-CD303 (clone 102G7; kindly provided by Dendritics, Lyon France), anti-MHC II (SLA-DQ, clone TH16B; VMRD), his-tagged porcine recombinant proteins Flt3L (14) and IL-3 (15), or RPE-conjugated anti-CD16 (clone G7; AdB Serotec). Anti-mouse IgG1-RPE, anti-mouse IgG2a-RPE (SBA) or anti-his-RPE (Miltenyi Biotec) were used as secondary Abs. For each marker, the corresponding FMO (fluorescence minus one) was used as negative control. All the FCM acquisitions were performed on a FACS Canto II (BD Bioscience) using the DIVA software and were further analyzed with the FlowJo software (TreeStar).

Prior to cell sorting of the different subsets identified previously, T cells were depleted from PBMCs. After incubation with anti-CD3 (hybridoma clone PTT3/FyH2; kindly given by Dr. K. Haverson, University of Bristol, U.K.), cells were incubated with anti-mouse IgG magnetic beads (Miltenyi Biotec) and sorted using an LD Column and the MACS system (Miltenyi Biotec). The CD3⁻ fraction was then stained with anti-CD172a, CD14, CD4 and CADM1 as described above, using Alexa Fluor 750-APC streptavidin instead of V500-conjugated streptavidin. CD14⁻CD172a^low CADM1⁺ putative cDC1, CD14⁻CD172a⁺CADM1⁺ putative cDC2, CD14⁻CD172⁺CADM1⁻CD4⁺ pDCs and CD14⁺ monocytes were then sorted using a FACS Aria (BD Bioscience) with the DIVA software.

For mRNA sequencing, total RNA was extracted from FACS-sorted cell subsets and in vitro-generated MDMs using TRIzol (Life Technologies) in combination with the Nucleospin RNA kit (Macherey Nagel). Briefly, cells were lysed with 1 ml TRIzol and kept at −70°C until further extraction. After thawing, 0.2 ml chloroform was added to the TRIzol lysate and the samples mixed vigorously and incubated for 3 min at room temperature. The extractions were then centrifuged at 12,000 × g for 15 min at +4°C. The aqueous phase was collected and mixed with 500 μl 75% ethanol and the RNA precipitated for 10 min at room temperature. The precipitate was resuspended gently and loaded on a NucleoSpin RNA column, and the RNA purified with the Nucleospin RNA kit including DNase treatment according to the manufacturer’s instructions. Finally, the RNA was eluted from the column with 40 μl RNase-free water. Quality and quantity of the purified RNA were assessed with an Agilent 2100 Bioanalyzer (Agilent Technologies) and a Qubit 2.0 Fluorometer (Life Technologies). Approximately 500 ng of high quality RNA (RNA integrity number RIN >8) was used for nondirectional paired-end mRNA library preparation (TruSeq Sample Preparation Kit; Illumina). Total mRNA libraries were randomly multiplexed in eight samples per lane and sequenced on the Illumina HiSeq2500 platform using 2 × 100 bp paired-end sequencing cycles. The Illumina BCL output files with base calls and qualities were converted into FASTQ file format and demultiplexed with CASAVA (v. 1.8.2) software. Except for the MDM RNAs, all samples were run on the same day.

Between 16.6 and 31.5 million read pairs were obtained per sample. The reads were mapped to the pig reference genome (Scrofa 10.2, Ensembl release 75) with Tophat v. 2.0.11 (16). Htseq-count v. 0.6.1 (17) was used to count the number of reads overlapping with each gene, as specified in the Ensembl annotation (release 75). The Bioconductor package DESeq2 v. 1.4.5 (18) was used to test for differential gene expression between two populations.

Heatmaps were built in R 3.2.2 to visualize the gene expression data and identify clusters of samples with similar expression profiles. The heatmaps were produced using the regularized log transformed counts (from DESeq2) for the N genes with the most variable expression across all samples, with different values of N between 500 and 5000. Agglomerative clustering with Euclidean distances and complete linkage was used. Similar tree topologies were obtained with different distance measures or linkage methods and hierarchical clustering. The R package pvclust v. 2.0-0 was used to assess the support of the observed clusters.

The RNAseq data are available in the European Nucleotide Archive (http://www.ebi.ac.uk/ena) under the accession number PRJEB15381.

Cell type specific gene expression signatures were compared between pigs, mice and humans using the conceptual approach outlined in Lamb et al. (19). Pig gene-expression signatures were obtained by identifying genes differentially expressed between each cell type and a pool of all other cell types using DESeq2 v. 1.4.5. The resulting gene lists were then sorted based on the false discovery rate–adjusted p values to have the most significantly upregulated genes at the top and the most significantly downregulated genes at the bottom. These ordered lists were compared with published microarray data from humans and mice [GEO datasets GSE35457 and GSE35458 from Haniffa et al. (20)].

The following cell types were selected for further analysis: human blood CD141 (cDC1), human blood CD1c (cDC2), human blood pDC, and human blood CD14 monocytes; mouse spleen CD8 cDC1, mouse spleen CD4 cDC2, mouse blood pDC, and mouse blood GR1 high monocytes. We used GEO2R (NCBI) to test for differential gene expression between each cell type and the remaining three cell types were pooled. Microarray probes were excluded from further analysis if they measured multiple genes or if the ortholog of the gene in pig could not be determined unambiguously. For genes measured by multiple probes on the array, only the one with the highest average expression level was retained [following Miller et al. (21)]. The final number of genes available for analysis was around 12,000 in both species. The gene lists from humans and mice were sorted in the same way as the RNA-seq based tables from pigs.

Each cell-type specific signature from pig was compared with all human and mouse signatures using the R package OrderedList v. 1.44.0 (22). This tool determines the number of shared elements in the first n elements of two lists and calculates a final similarity score where genes receive more weight the closer they are to the top or bottom of the list. This ensures the score is dominated by the genes showing the most significant differential expression. We report similarity scores for n = 1000. The relative similarity among the cell types was generally consistent for other values of n (assessed for values between 100 and 2500). To assess the statistical significance of the similarity scores, the observed values were compared with a null distribution obtained by reshuffling the genes. Because invariant genes do not influence the similarity score, the middle 60% of genes were excluded from the permutations.

To assess the cell-surface costimulatory molecule expression, PBMCs, IL3R-depleted or IL-3R-reconstituted PBMCs were plated in 12-well plates at 5 × 10⁶ cells/well in 1 ml DMEM (Life Technologies) supplemented with 10% FBS (Biochrom), 1% nonessential amino acids (Life Technologies), 1% sodium pyruvate (Life Technologies), 1% Hepes (Life Technologies), and 20 μM 2-ME (Invitrogen). For IL-3R depletion, PBMCs were first incubated with a his-tagged recombinant porcine IL-3 (15). Cells were then incubated with an anti-his Ab (Roche), and with anti-mouse IgG magnetic beads (Miltenyi Biotec), before being sorted using an LD Column and the MACS system (Miltenyi Biotec). The PBMCs, IL-3R⁻ fraction and IL-3R⁺ fraction were kept on ice overnight before being plated for stimulation. For the IL-3R–reconstituted PBMCs, the IL-3R⁺ fraction was added to the IL-3R⁻ fraction before plating the cells. Cells were stimulated 5 h with the following TLR ligands: 10 μg/ml PAM₃Cys-SKKKK (PAM₃Cys L2000; EMC microcollections), 10 μg/ml polyinosinic-polycytidylic acid (poly I:C; Sigma), 1 μg/ml LPS (Escherichia coli O111:B4; Sigma), 5 μg/ml gardiquimod (InvivoGen), 5 μg/ml CpG oligodeoxynucleotide (CpG, sequence D32, 5′-ggTGCGTCGACGCAGggggg-3′; Eurofins genomics) or left unstimulated (controls). Cells were then harvested and the DC subset staining was performed as described above, using anti-human CD40 (hybridoma clone G28-5, ATCC HB-9110) or the human CD152-μ Ig (human CTLA fusion protein; Ancell) and corresponding secondary Abs.

PBMCs were plated in 12-well plates at 5 × 10⁶ cells/well in 1 ml supplemented DMEM. For TNF-α intracellular staining, cells were stimulated 1 h with the different TLR ligands (PAM₃Cys, poly I:C, LPS, gardiquimod, CpG) or left unstimulated as controls. Brefeldin A (eBioscience) was then added to the culture for 4 h to stop the cytokine secretion. For IFN-α intracellular staining, PBMCs were stimulated 3 h before adding brefeldin A for an extra 4 h. Cells were then harvested and fixed in 4% paraformaldehyde. After a wash with 0.1% saponin (Panreac Applichem), cells were incubated with either Alexa Fluor 647-conjugated anti-human TNF-α (clone Mab11; Biolegend), or anti-porcine IFN-α (clone F17; PBL Assay Science) in 0.3% saponin. For the IFN-α staining, cells were then incubated with anti-mouse IgG1 RPE (SBA) in 0.3% saponin after another 0.1% saponin wash.

FACS-sorted cell subsets (cDC1, cDC2 and pDCs) were plated in 96-well plates with 5 × 10⁴ cells/well in a final volume of 100 μl of complemented DMEM and stimulated with either poly I:C, LPS, gardiquimod, CpG or left unstimulated, followed by culture at 39°C with 5% CO₂ for 18 h. The supernatants were harvested and cytokine concentrations were assessed using the porcine ProcartaPlex multiplex Luminex xMAP assay (eBiosciences) according to the manufacturer’s instructions and read on a MAGPIX machine (Bio-Rad).

FACS-sorted cell subsets (cDC1, cDC2, and pDCs) were plated in 96-well plates with 5 × 10⁴ cells/well in a final volume of 100 μl of complemented DMEM and stimulated with either PAM₃Cys 10 μg/ml, CpG 5 μg/ml or left unstimulated as control, followed by culture at 39°C with 5% CO₂ for 3 h. Cells were then harvested in 1 ml TRIzol and kept at −70°C until mRNA extraction. mRNA was extracted using the Nucleospin RNA kit (Macherey Nagel) as described above. Reverse transcription and quantitative PCR (RT-QPCR) was realized in two steps. The reverse transcription step was performed using the Omniscript RT kit (Qiagen) following the manufacturer’s instructions, and random primers 3 μg/μl (Invitrogen). Generated cDNA was then used to perform QPCR with the Taqman Fast Universal Master Mix 2×, No AmpErase (Applied Biosystems) using the ABI PRISM 7700 sequence detector system (Applied Biosystems). The relative expression of IL-12p35, IL-12p40, and IL-23p19 was calculated by the comparative cycle threshold method using the housekeeping gene 18S as reference genes to normalize the data (23). Primers and probes used for QPCR were the following: 18S forward primer 5′-CGC CGC TAG AGG TGA AAT TC-3′; 18S reverse primer 5′-GGC AAA TGC TTT CGC TCT G-3′; 18S probe 5′-TGG ACC GGC GCA AGA CGG A-3′; IL-12p35 forward primer 5′-GCC CAG GAA TGT TCA AAT GC-3′; IL-12p35 reverse primer 5′-GGG TTT GTT TGG CCT TCT GA-3′; IL-12p35 probe 5′-CAA CCA CTC CCA AAA TCT GCT GAA GGC-3′; IL-12p40 forward primer 5′-TCT TGG GAG GGT CTG GTT TG-3′; IL-12p40 reverse primer 5′-AAG CTG TTC ACA AGC TCA AGT ATG A-3′; IL-12p40 probe 5′-ACC AGC AGC TTC TTC ATC AGG GAC ATC A-3′; IL-23p19 forward primer 5′-CAG AAG AGG GAG ATG ATG AGA CTA CA-3′; IL-23p19 reverse primer 5′-GGT GGA TCC TTT GCA AGC A-3′; and IL-23p19 probe 5′-CTG AGG ATC ACA GCC ATC CCC GC-3′.

For FCM costimulatory molecule expression and intracellular cytokine production, statistical significance was assessed using the unpaired nonparametric Kolmogorov–Smirnov test (**p ≤ 0.01, *p ≤ 0.05). All statistics were calculated using the GraphPad Prism 6 software.

To identify candidates for porcine cDC1, cDC2, and pDC, we combined CD14, CADM-1, CD172a, and CD4 to define clearly distinct subsets of mononuclear cells. After doublet discrimination, we gated on cells with a bigger size (high forward scatter) and higher granularity (high side scatter) to exclude most of the lymphocytes. CD14 expression was then used to define monocytes. Within the CD14⁻ cells, we identified three subpopulations of interest: a CD4⁻CD172a^lowCADM1⁺ subset, a CD4⁻CD172a^highCADM1⁺ subset, and a CD4⁺CD172a⁺CADM1⁻ subset (Fig. 1A, Supplemental Fig. 1A). This gating strategy allowed the definition of clearly distinct subsets, which expressed the conserved DC marker Flt3 (CD135) (Fig. 1B) supporting the hypothesis that these would be bona fide DC. It was clear from previous work that although all blood DCs expressed Flt3, the CD4⁺ subset represented pDC (8). Again, based on published work (24), we hypothesize that the CD4⁻CD172a^lowCADM1⁺ subset represents the cDC1. The absence of CD14 together with a high expression of CD172a and the expression of Flt3 made the CD4⁻CD172a^highCADM1⁺ subset an interesting candidate as equivalent of cDC2. To further evaluate this hypothesis, we analyzed a series of markers by flow cytometry. The M-CSF receptor (CD115) was highest on monocytes, followed by the putative cDC2, but was not found on putative cDC1 or on pDCs (Fig. 1B). The expression of wCD11R1, recognizing the same molecule as the human cross-reactive anti-CD11b, was found on putative cDC1 and cDC2, but not on pDCs and only weakly on a subset of monocytes (Fig. 1B). CD1a was found to be highly expressed on the putative cDC2 subset, whereas very low to no expression was observed on cDC1 and pDCs (Fig. 1B). The proposed cDC1 subset displayed the strongest expression of CD205, a receptor recognizing apoptotic and necrotic self, but the molecule was also found on other mononuclear cells (Fig. 1B). Only pDCs expressed IL-3 receptor (CD123) and CD303, the latter being identified with an anti-human cross-reactive Ab, confirming their correct definition (Fig. 1B). We observed a low surface expression of the scavenger receptor CD163 on the cDC2 subset, and no expression on cDC1 or pDCs (Fig. 1B). The FcγRIII receptor CD16 was surprisingly widely expressed, with the highest expression found in monocytes and cDC2, whereas expression was weaker in pDCs (Fig. 1B). In conclusion, the use of mAbs against CD1a, CD4, wCD11R1, CD14, CD115, CD123, CD135, CD163, CD172a, CD303, and CADM1 permits an unambiguous definition of clear subsets suitable to transcriptional analysis.

From the blood of three different animals, we sorted the four populations described in Fig. 1 and applied mRNA sequencing. Principal component analysis (PCA) of the 500 most variable genes from the mRNA sequencing data from these samples confirmed that we had sorted clearly distinct populations (Fig. 2A). This was supported by a heat map analysis representing the expression of the 500 most variable genes using the Euclidian distance with complete linkage method (Fig. 2B). Indeed, for each cell subset, cells from the three different animals clustered together, confirming that the transcription profiles of these subsets were similar. As expected, the two cDC subsets clustered together, but unlike what was described for mice and humans, we found the monocytes to be closer to the cDC subsets than the pDCs (Fig. 2B). These results were confirmed using a higher number of genes for the analysis (1000 and 5000 most variable genes, Supplemental Fig. 2), but also with additional methods of clustering analysis (Euclidian distance, average or single linkage methods, Supplemental Fig. 3), and all nodes were supported by high bootstrap values (≥92%; Supplemental Fig. 4). The study of the morphology of the cells confirmed the sorted cells were homogeneous and displayed the characteristics of cDC1, cDC2, and pDC subsets described in other species; also, they were distinct from monocytes (Supplemental Fig 1B).

The RNAseq data set obtained for each porcine putative subset was used to compare the gene expression signature of the porcine cell subsets to their putative human and murine counterparts, making use of published microarray data (20). A similarity score was obtained by comparing the cell type–specific signature to each of the other subsets. Porcine cDC1, pDC, and monocyte signatures were most similar to the signature of their human and mouse counterparts (all similarity scores with empirical p value <0.001; Fig. 2C). Although the porcine cDC2 signature was similar to the human cDC2, no clear pattern was found when compared with the mouse subsets; however, this was also not found when comparing the human cDC2 subset to its mouse counterpart. These results support our identification of the porcine cDC1, cDC2, and pDC subsets, and that their gene expression signature for cDC2 is closer to their human counterpart.

Using the RNA sequencing data, we then characterized gene expression for the putative DC subsets, first focusing on subset-specific genes. We only considered the genes that were significantly upregulated at least 2-fold compared with all the other subsets, with a lower limit cut-off of 250 reads. Of the three DC subsets, with 584 overexpressed genes, pDCs were the subset displaying the most differentially upregulated genes compared with the putative cDC1, putative cDC2, and monocytes. In the cDC1 subset we identified 172 specifically upregulated genes and 44 genes in cDC2. To validate the mRNA sequencing data, we focused on the number of reads obtained for the markers for which we had determined the cell surface expression by FCM. For all the genes with an identified sequence (CD14, CADM1, CD4, FLT3, CSF1R, ITGAM, PCD1A, and LY75), the mRNA expression levels related to protein expression determined by FCM in terms of subset distribution (Fig. 3). This validation of the mRNA sequencing data allowed us to further investigate the specific gene expression of each of the subsets identified.

Table I shows selected genes that were upregulated in all putative DC subsets (cDC1, cDC2, and pDCs) compared with monocytes and MDMs, as well as genes that were specifically overexpressed by the two putative cDC subsets. As expected, the cross-species panDC markers FLT3 and BCL11A were found to be mostly DC specific. VCL (vinculin), one of the components of the podosomes, structures involved in Ag sampling in DCs (25), was also upregulated by all porcine DC subsets.

Concerning cDC shared markers, we identified CD11c (ITGAX) and CD2 as being specifically expressed by porcine cDCs. PAK1 coding for an important element of the immunological synapse (26) was overexpressed in the cDC. Two genes involved in TLR signaling, either as part of the signaling cascade (IRAK2) (27) or as inhibitor (SIGIRR) (28), were also strongly expressed by cDCs compared with the other cell populations. Two other genes that have already been reported as being upregulated in murine (NAPSA) (29) or porcine (NAV1) (9) cDCs were also overexpressed in our cDC subsets.

Selected genes overexpressed in porcine pDC are displayed in Table II. These include the two known transcription factors TCF4 (E2-2) and RUNX2, important for pDC development (4), CCR5, CD8B, PLAC8, and NRP1 (coding for CD304 or BDCA-2), the latter being used to identify human pDCs. Surprisingly, the scavenger receptor CD36, and CLEC12A, an anti-inflammatory receptor for dead cells (30), were also highly expressed by pDCs. UNC93B1, coding for a protein responsible for the delivery of TLR3, TLR7, and TLR9 to endolysosomes (31) was also found to be more expressed in pDCs. We also found the lysosomal marker LAMP3 (DC-LAMP, CD208) and the notch signaling pathway receptor NOTCH3 to be highly overexpressed in pDCs. To our surprise, porcine pDCs specifically overexpressed the gene coding for TNF-α and other TNF-associated genes such as TNFAIP8, TIAF1, and TRAF4, although the cells were sorted in a resting state. As previously described in pigs (9), pDCs were overexpressing the genes coding for the BCR signaling-related molecule BLNK, the transcription factor XBP1, important for mouse DC ontogeny (32), and the surface receptor LRP8, which has a function that remains unknown. Strikingly, several genes related to the complement system (C2, C3, C5, CD93) were also overexpressed in porcine pDCs.

Table III shows a selection of genes specifically overexpressed in the putative cDC1. XCR1, BATF3, APN (CD13), and DPP4 (CD26) were strongly and mostly exclusively expressed in this subset, confirming the proposed cDC1 identity in agreement with previous studies (9–11). In addition, SCIMP, coding for a protein involved in MHC class II signaling (33), IL21R, CD226, CD34, WDFY4, and GCSAM were expressed almost exclusively in cDC1. The transcription factor BCL6 was also found at intermediate levels in the putative cDC2, monocytes or MDMs. The highest levels of TAP1, TAP2, and PSMB8, all genes linked to MHC class I Ag processing, support this functional specialization also in porcine cDC1. The C-type lectin receptor CLEC7A, also known as dectin-1, which recognizes fungal β-glucans and is involved in anti-fungal immune responses, was also overexpressed in the porcine cDC1 cells.

Table IV shows genes differentially regulated in the putative cDC2 subset. Similar to the human cDC2 counterpart (34), this subset expresses high levels of FCGR2B (CD32), FCER1A, and MRC1 (CD206). Furthermore, two members of the NOTCH signaling pathway, NOTCH4 and RBPJ, were found to be overexpressed in the porcine cDC2 subset. The fact that RBPJ-deficient mice were specifically depleted for cDC2 (35) further supports the correct identification of this subset. Furthermore, the expression of ZEB2 and the lack of IRF8 expression also matched what was observed in mouse cDC2 (36).

Table V displays the expression of known monocyte and MDM-specific genes. As shown in Fig. 1 by FCM, CD14 was expressed very specifically in monocytes, as was SIGLEC1 (CD169), with almost no expression of those genes in any of the DC subsets. The macrophage-specific genes CD68, CD109, CD209, and MMP9 were expressed exclusively in MDMs and absent from all the DC subsets, except for a very low expression of CD68 in the cDC subsets.

Together, our phenotypic data and the unique expression of conserved markers specific for cDC1, cDC2, pDC, and monocytes in our defined subsets provide strong evidence that these represent true and pure cDC1, cDC2, and pDCs subsets in pigs. Nevertheless, our data showed also some particularities of the porcine MPS.

CD40 mRNA was expressed at high levels by both cDC subsets and to a lesser extent by monocytes (Fig. 4A). Both cDC subsets expressed also the CD80, CD83, and CD86 genes, whereas, surprisingly, monocytes showed the highest expression of CD80 and CD86, and a strong expression of CD83. pDCs expressed very low levels of CD40, CD80, and CD83, whereas the levels of CD86 gene expression were comparable to those in cDC (Fig. 4A). FCM demonstrated that both, the cDC1 and the cDC2 subsets expressed very high levels of SLA-DQ MHC class II on their cell surface. Also, the majority of monocytes were positive, whereas pDCs displayed a more variable expression profile (Fig. 4B). The mRNA sequencing data confirmed this very high expression of many SLAs by the cDC1 and cDC2 subsets, with an intermediate expression in monocytes and the lowest expression in pDCs and MDMs (Fig. 4C). This expression pattern was also found for CIITA, a molecule that acts as a transactivator of the MHC class II gene expression.

We next used the RNA sequencing data to investigate the expression of a broad range of PRRs in the defined MPS subsets (Fig. 5). The expression levels of TLR1, TLR2, TLR4, TLR6, and TLR8 were the highest in monocytes, the lowest in cDC1 and pDC, and intermediate in cDC2 (Fig. 5). As described in humans and mice, porcine pDCs expressed the highest levels of TLR7 and TLR9 (Fig. 5), but in pigs, TLR7 was also expressed by monocytes and the cDC2 subset, whereas the cDC1 subset expressed high levels of TLR9. Another peculiarity in pigs was the restricted expression of TLR3 in pDCs (Fig. 5), whereas in mice and humans, this receptor was found to be expressed in the cDC1 subset (37, 38). The NOD-like receptor (NLR) NOD1 was mainly restricted to cDC2 and monocytes (Fig. 5). The inflammasome-related NLRP3 was highly expressed by monocytes and cDCs (Fig. 5). The two RIG-I like receptors IFIH1 (MDA-5), and DHX58 (LGP2) displayed a similar pattern of expression with high expression in monocytes, lower expression in both the cDC2 and pDC subsets and the lowest expression in the cDC1 subset.

Considering the above expression profile of PRR, the response of each MPS subset to a broad range of TLR ligands was assessed in terms of cytokine secretion and costimulatory molecule expression. To this end, PBMCs were stimulated with TLR ligands triggering TLR2/6 (PAM₃Cys), TLR3 (poly I:C), TLR4 (LPS), TLR7 (gardiquimod), and TLR9 (CpG D32), or left unstimulated as a control. The ability of DC subsets to secrete or express key cytokines in response to TLR ligands was examined on sorted blood MPS subsets. The limited number of cells obtained postsorting did not allow us to test the whole TLR ligand panel on all the subsets. The production of nine different cytokines was determined using a multiplex immunoassay. Regardless of the DC subset and the TLR stimulation, low levels of IFN-γ (<90 pg/ml), IL-1β (<80 pg/ml), IL-10 (<60 pg/ml), IL-4 (<10 pg/ml), or IL-6 (<30 pg/ml) were detected. Surprisingly, increases in TNF-α, IL-12p40, and IFN-α production were only detected in the supernatants of pDC cultures and were mostly induced by CpG, to a lesser extent by gardiquimod and only occasionally by poly I:C (Fig. 6A). CXCL8 and TNF-α were the only cytokines that were produced by a DC subset other than pDCs. Indeed, cDC1 and cDC2 cells were all able to produce CXCL8: poly I:C and CpG stimulations did not influence the cDC1 production of CXCL8, whereas an increase was observed for cDC2 following poly I:C or LPS stimulation (Fig. 6A). Given the lack of immunoassay to differentiate IL-12 and IL-23, we also assessed the mRNA expression of IL-12p35 (specific subunit of IL-12), IL-23p19 (specific subunit of IL-23), and IL-12p40 (common subunit of IL-12 and IL-23) by the different subsets following PAM₃Cys and CpG stimulation (Fig. 6B). IL-23p19 was consistently induced in cDC2 by PAM₃Cys, although individual samples from the other cell subsets also had comparable levels of mRNA (Fig. 6B). For IL-12p35, no expression was found in cDC1, only low levels in one of three animals in cDC2, and high levels in pDCs after stimulation of CpG. A similar pattern was found for IL-12p40 with the highest levels of mRNA found in pDC after CpG stimulation. Stimulation with PAM₃Cys also consistently induced IL-12p40 expression by the pDC (Fig. 6B). These results confirm the results with the immunoassay (Fig. 6A) indicating that within the peripheral blood DCs, the pDCs provide the Th1-polarizing cytokine IL-12.

To rule out that the inconsistency in or lack of cytokine responses were caused by a negative impact of the cell sorting process on cell functionality, TLR ligand stimulation was tested with PBMC using intracellular cytokine detection by multicolor FCM. PBMCs were stimulated with the same TLR ligands used previously for either 5 h (for TNF-α) or 7 h (for IFN-α). The percentage of cells expressing the cytokines as well as the mean fluorescence intensities (MFI, reflecting the amount of cytokine) were determined by FCM. Similar to what was observed with sorted cells, pDCs were the main producers of both TNF-α and IFN-α (Fig. 7). CpG, gardiquimod, PAM₃Cys, and poly I:C induced a high percentage of TNF-α–producing pDCs and a significant increase in MFI, gardiquimod being the most efficient (Fig. 7A). LPS only induced TNF-α in a low percentage of pDC, with MFIs close to those of unstimulated controls (Fig. 7A). No TNF-α induction was observed in cDC1 following stimulation with any of the TLR ligands (Fig. 7A). A significant but very low (<5%) increase in TNF-α–producing cells was found in cDC2 following PAM₃Cys and CpG stimulation, and some samples did not respond. TNF-α was also induced in monocytes especially following PAM₃Cys (<12%) and LPS (<18.5%) stimulation, but compared with pDC the MFIs were low (Fig. 7A). As expected, pDCs were also found to be the main IFN-α producers, with CpG the most efficient stimulus (between 40 and 87.9% of positive cells) followed by gardiquimod (2.7–15.9% of positive cells) (Fig. 7B). Statistical significance was also found in terms of IFN-α positive cells following CpG stimulation for the cDC1 and cDC2 subsets, but the percentages and MFIs were very low and not all samples reacted (Fig. 7B). PAM₃Cys, LPS, gardiquimod, and CpG did also statistically increase the percentage of IFN-α positive monocytes (Fig. 7B), but again the percentage remained around 1% and this did not always translate to an increase in MFI (Fig. 7B).

We then assessed the ability of these different MPS subsets to upregulate their surface expression of costimulatory molecules following TLR stimulation. When left unstimulated, cell-surface expression of CD40 and CD80/86 by pDCs was lower compared with cDC1, cDC2 or monocytes (Fig. 8A), which was in accordance with the transcriptomic data shown in Fig. 4. All TLR ligands induced a significant increase in the cell-surface expression of CD40 as well as CD80/86 in the cDC1 and the cDC2 subsets (Fig. 8A). A lower but significant upregulation of surface expression of CD40 and CD80/86 was also observed for pDCs following PAM₃Cys, poly I:C, gardiquimod, and CpG stimulation. LPS did not have any effect on the levels of CD40 in pDC and induced a low but significant increase of CD80/86 (Fig. 8A). In monocytes, two different pictures were observed with TLR stimulation. Although all TLR ligands induced a low but significant increase of CD40, no increase was observed in terms of CD80/86 expression. The upregulation of costimulatory molecules following TLR ligand stimulation did not entirely match the TLR expression patterns described in Fig. 5. Because pDCs were found to produce high amounts of TNF-α and IFN-α, we wanted to assess a potential indirect activation of the other subsets by these cells. To this end, we stimulated the IL-3R⁻ pDC–depleted fraction of PBMCs with the same broad panel of TLR ligands and measured the expression of CD40 and CD80/86 by the three remaining MPS subsets (monocytes, cDC1, and cDC2). The cDC1 subset displayed a significant increase in both CD40 and CD80/86 surface expression following PAM₃Cys and poly I:C stimulation (Fig. 8B). These two TLR ligands also induced a slight but significant increase in CD40 expression by the monocytes, whereas poly I:C surprisingly induced a downregulation of CD80/86. Another striking feature was the absence of CD40 upregulation by cDC2 following any of the TLR ligand stimulation. Only poly I:C for CD40 and poly I:C and LPS for CD80/86 induced a slight but significant increase (Fig. 8B). In the absence of pDCs, CpG stimulation did not induce any upregulation of costimulatory molecules, and gardiquimod surprisingly induced significant downregulation of CD80/86 in all three subsets (Fig. 8B). When the IL-3R⁺ fraction was reconstituted, we observed CD40 and CD80/86 expression profiles following TLR ligand stimulation similar to what was observed with whole PBMCs (Fig. 8C). A significant increase of CD40 and CD80/86 in cDC1 and cDC2 subsets followed CpG stimulation. CD40, but not CD80/86, was also increased in the cDC1 subset after reconstitution. Upregulation of CD40 and CD80/86 by cDC2 following PAM₃Cys stimulation was also restored. Thus, pDCs seem to play an important role in cDC2 subset activation and appear to be required for the upregulation of costimulatory molecules by APCs following TLR7 or TLR9 stimulation.

Only a few studies have started to decipher porcine MPS in the blood (9, 14, 15), skin (10), or lungs (11) using cell sorting, microarrays, or RT-QPCR. Here, we combined an extensive phenotypic description permitting sorting of pure cell subsets followed by mRNA sequencing, to enable an unbiased and comprehensive transcriptional profiling of porcine blood DC subsets and monocytes.

Like others we found cross-species similarities, which were used to identify the subsets. Phenotypic characterization of the cDC1 subset showed species-conserved features such as high surface expression of CD135, CADM1, CD205, low levels of CD172a, and a lack of CD115. XCR1, BATF3, and APN gene expression were highly restricted to this subset, as described by other studies (9–11), confirming the correct definition of cDC1. The high levels of MHC class I Ag processing-related gene expression such as TAP1, TAP2, or PSMB8 in porcine cDC1 would indicate that in pigs this DC subset has a similar functional specialization as in other species. Porcine blood cDC1 also displayed differences when compared with human or mouse counterparts, such as the specific expression of CLEC7A, which is expressed by both human cDC subsets (34), or low expression of NFIL3, a transcription factor important for the development of the mouse cDC1 subset (39). On porcine MPS cells, NFIL3 was largely overexpressed by monocytes. CD11b (equivalent to porcine CD11R1) was highly expressed on cDC1 and cDC2, contrasting with murine DC, where CD11b is used as a cDC2-specific marker.

This study allowed an extensive phenotypic and functional characterization of the porcine blood cDC2 subset. The expression of certain monocyte markers such as CSF1R and CD163 on cDC2 reflects the difficulty in defining this subset, but the absence of the porcine monocyte–specific markers such as CD14 and SIGLEC1 combined with a high expression of FLT3, BCL11A, ITGAX, PAK1, NAPSA and SLA-DR, SLA-DQ, and CIITA confirmed these as DCs. This was also supported by the clustering and PCA analysis based on transcriptome data demonstrating a close relationship to cDC1. Furthermore, porcine cDC2 shared markers with human cDC2 (34) including FCGR2B (CD32), FCER1A, and MRC1 (CD206). Porcine cDC2 also display species-specific characteristics with high surface expression of CD1.1, the porcine equivalent of CD1A (40), which is only found in human dermal but not blood cDC2 (41), and CADM1, which is restricted to the cDC1 subset in other mammalian species (42).

The porcine pDC subset had already been characterized and extensively studied in pigs focusing on its great ability to produce IFN-α in response to viruses (15, 43–48), but our results also describe similarities between porcine and human pDC such as the expression of CD303 (BDCA-2), CD304 (NRP1 or BDCA-4), BLNK, and PLAC8 (49–51). Despite this, we also observed several discrepancies such as the high expression of CD36 or CLEC12A, which have been shown to be more broadly expressed in human DC subsets and macrophages (52, 53). As in mice (29) but not in humans (54), NOTCH3 was also upregulated in porcine pDCs. A remarkable peculiarity was the exclusive and high expression of complement-related genes (C2, C3, C5, and CD93) by porcine pDCs, whereas these are expressed in several human DC subsets (55). This would suggest a more prominent role of the porcine pDCs in complement biology compared with other species.

The expression of PRRs by the different porcine blood DC subsets displayed similarities with other species such as the highest levels of TLR7 and TLR9 expression by porcine pDCs, but showed also some species-specific features. Indeed, porcine pDCs displayed the highest levels of TLR3, which in mice and humans is restricted to the cDC1 subset (37, 38). Unlike in human DCs, TLR9 was also found at relatively high levels in porcine cDC1, more resembling the murine distribution (37).

Surprisingly, cytokine production following TLR ligand stimulation was restricted almost exclusively to pDCs. Sorted pDCs were the only subset to produce high amounts of IFN-α, TNF-α, and IL-12p40 especially following CpG (TLR9), gardiquimod (TLR7), and, to a lesser extent, poly I:C (TLR3) stimulation. pDCs were also the only subset to highly express IL-12p35 following CpG stimulation, which would make these cells an important source of IL-12 in pigs. None of the porcine cDC subsets were able to produce any IL-12, although these two subsets are the main source of this cytokine in humans (56, 57). Because of the small numbers of cells that were sorted, not all conditions could be tested for each subset, especially for the cDC1 subset, which is scarce in blood. Moreover, we could not rule out an effect of the sorting process on the functionality of these subsets. Therefore, we assessed their ability to produce TNF-α and IFN-α by intracellular FCM, confirming the results we obtained in multiplex immunoassay with the sorted subsets. pDCs were also unexpectedly responsive to the TLR2 ligand PAM₃Cys in terms of TNF-α production. It appeared again that TLR ligand responses in terms of cytokine secretion did not always correlate with the TLR expression levels, indicating that other elements are also involved in determining this. When using PBMC, this differential distribution of TLR did not affect the different DC subset responses to a broad range of TLR ligands, with the two cDC subsets significantly upregulating costimulatory molecules (CD40 and CD80/86) following TLR2, 3, 4, 7 and 9 stimulation. However, in pDC-depleted cultures no costimulatory molecule upregulation by the two cDC subsets following stimulation with TLR7 and TLR9 was observed. The cDC2 subset was the most dependent on this indirect activation by pDC and was also unresponsive to PAM₃Cys and less responsive to LPS. Interestingly, the reconstitution of the pDC fraction fully restored the cDC1 and cDC2 responsiveness to CpG and the cDC2 responsiveness to PAM₃Cys. CD40 upregulation by cDC1 following TLR7 stimulation was also restored. Altogether, these results would imply a split of functions in the porcine DC system, with cDC1 and cDC2 being very efficient in presenting Ag whereas pDCs appear to have a critical role in cytokine production providing “signal 3” for induction of naive T cell responses. Similar observations were made with porcine monocyte–derived DCs shown to depend on pDC-derived cytokines for maturation (58). Future studies are required to address this question in vivo, but there is currently no porcine pDC-depleted model available.

Most blood DC originate from the bone marrow and will eventually migrate into both nonlymphoid and lymphoid tissues, where they will be influenced by the local environment. From mouse and human studies, it is known that this can result in tissue-specific phenotypic and functional changes (4). Although we have preliminary data indicating that DC subsets can also be identified in tissues using the described markers, future studies focusing on primary and secondary lymphoid tissue as well as nonlymphoid tissue such as the skin, the mucosal lamina propria and the organ parenchyma are required to confirm DC subset identity. In addition, tissue-specific transcriptomic signatures will help to generate a more complete picture of the porcine DC system. Considering that tissue DC originate from blood DC, our data represent an important basis for such studies.

The present study also adds important data relevant for understanding porcine DC and MPS functions, which are useful for studies on host-pathogen interactions and for the rational design of powerful vaccines such as those targeting DC. To this end, a wide variety of receptors have been targeted in pigs, including MHC molecules, siglecs, TLRs, C-type lectin receptors, costimulatory molecules, and chemokine receptors, without clear information on receptor distribution (59). Furthermore, even if the distribution of targeted receptors is conserved between species, other peculiarities in the biology of the DC subset could modify the outcome of Ag delivery. The same holds true for targeting PRRs to enhance immune responses. For example, targeting TLR3 in pigs would be directed toward pDC, not cDC1.

These interspecies differences found in terms of gene expression and functions among the DC subsets reflect evolutionary disparities, with animals evolving in different environments. Indeed, exposure to different pathogens, different food habits, and digestive tracts will have an influence on shaping the immune system, leading to distinct species-specific differences. There are many other examples of peculiarities of the porcine immune system, which have been reviewed recently (60). Although the mouse model is very useful in the discovery and study of the main functions of the immune system, extrapolation of the findings to other species requires thorough immunological investigation in the target species to successfully translate immunological knowledge into vaccines and immunotherapeutics.

We thank the Next Generation Sequencing Platform of the University of Bern, in particular Muriel Fragnière, for performing the porcine macrophage mRNA library preparation and sequencing. We thank also the animal caretakers Hans-Peter Lüthi and Andreas Michel for blood sampling.

The authors have no financial conflicts of interest.