The transcription factor KLF4 reconstitutes macrophage differentiation from PU.1^-/- fetal liver cells (30) and is essential for murine monocyte differentiation in vivo (31). In human immunohistology, KLF4 exceeds other monocyte-affiliated markers in monocyte lineage assignment (32) and induces monocyte markers CD14 and CD11b when transfected into myeloid cell lines (30). Moreover, KLF4 interferes with TGF-β1-mediated impairment of pro-inflammatory cytokine production by monocytic cells (33). We recently demonstrated that Notch represses KLF4 in human CD14⁺ monocytes, thereby rendering these cells capable of differentiating into LCs in response to TGF-β1/RUNX3 signaling. Mechanistically, KLF4 bound to the RUNX3 promoter in KLF4^hi monocyte-derived cells, likely interfering with RUNX3 transcription (34). Since in contrast to monocytes, CD1c⁺ blood cDC2 do not rely on Notch co-stimulation for differentiating into LCs in vitro (21), we assessed their KLF4 protein expression status. We purified CD1c⁺ blood cDC2s from mononuclear cell preparations by first depleting CD14⁺ monocytes and CD19⁺ B cells, followed by CD1c positive selection. While expectedly KLF4 is expressed by virtually all CD14⁺ monocytes, cDC2s lack KLF4 (Figure 1A). To address whether absence of KLF4 in cDC2 is associated with augmented TGF-β1/Notch signaling, we subsequently analyzed active Notch-1 (aN1) and canonical phospho-SMAD2/3 (p-SMAD2/3). Immunocytology analysis revealed that blood cDC2s exhibit elevated aN1 and p-SMAD2/3 levels relative to CD14⁺ monocytes (Figure 1B). Intracellular flow cytometry, albeit less sensitive, were supportive of these findings (Supplementary Figure 1). To validate these observations and gain mechanistic insights, we performed integrative data analysis of RNA seq datasets from blood cDC2s and CD14⁺ monocytes. We asked whether the increased activation of Notch and TGF-β signaling pathways in blood cDC2s can be verified based on their transcriptional profile. We therefore dissected the Upstream Regulators attributed to the blood cDC2s and CD14⁺ monocyte associated signature genes [based on (10)] using the Ingenuity Pathway Analysis platform (IPA). Among the significant Upstream Regulators, we identified TGF-β and Notch- attributed gene sets (defined by IPA and named therein as groups) for both cell types. Although the Notch and TGF-β group were similarly ranked in the blood cDC2s relative to CD14⁺ monocytes (Notch: p-value =1.53E-02, position 289/527 versus p-value =4.64E-04, position 466/2063; TGF- β: p-value = 2.21E-02, position 347/527 versus p-value =1.69E-02, position 1303/2063), the Notch group was predicted to be inhibited in CD14⁺ classical monocytes. Consistent with our immunostaining results (Figure 1A), we identified KLF4 only in CD14⁺ classical monocytes with an activated predicted state (p-value = 3.18E-03; position 725/2063) (Figure 1C and Supplementary Table 4). The analysis of known TGF-β regulated molecules in the monocyte/DC system also showed consistent results. Members of the TAM family of efferocytosis receptors are inversely controlled by TGF-β1 in cells of the mononuclear phagocyte system, with Axl and its ligand Gas6 induced, and MerTK repressed (29). Analyzing datasets from four different transcriptomic studies revealed that Axl/Gas6 are indeed differentially expressed by peripheral blood cDCs versus classical CD14⁺ monocytes, with cDCs showing upregulated expression of Axl and ligand Gas6. In contrast, TAM receptor MerTK exhibited reduced expression levels in cDCs compared to CD14⁺ monocytes. Expectedly, Tyro3 levels remained similar in both cell types (Figure 1D) (29). In conclusion, cDC2s lack detectable KLF4 protein, and show elevated constitutive Notch and TGF-β signaling relative to CD14⁺ monocytes.

Since we previously observed that KLF4 is upregulated upon moDC and macrophage differentiation of CD14⁺ monocytes in response to GM-CSF/IL-4 and M-CSF/IL-6, respectively (34), we asked whether KLF4 can be induced in blood cDC2s in response to these stimuli. Indeed, we observed that cDC2 can be induced to acquire moDC (CD11b⁺CD1a⁺CD209⁺) (Figure 2A) and macrophage (CD14⁺CD11b⁺CD206⁺) (Figure 2B) phenotypic characteristics in response to GM-CSF/IL-4 and M-CSF/IL-6 stimulation, and this is accompanied by KLF4 induction, with KLF4 expression levels equalling or exceeding those observed in monocytes cultured in parallel under identical cytokine conditions (Figure 2C). Expectedly (34), CD1c⁺ cell-derived CD207⁺ LCs generated in response to GM-CSF plus TGF-β1 lacked any detectable KLF4 protein (Figure 2C). Moreover, pooled (n=6) sub-sorted fractions of blood cDC2s (i.e. CD5⁺ versus CD5^lo/- (9), exhibited similar potency to differentiate into macrophages and moDCs, with the limitation that CD5⁺ cells failed to gain high levels of CD209, a key marker for moDCs (35) that is selectively regulated in moDCs by activation signals (36) (exemplified in Figure 2D). Therefore, blood cDC2s can be induced to differentiate into KLF4^hi cells exhibiting phenotypic characteristics of moDCs and macrophages.

The enlarged lesional psoriatic epidermis is populated by at least three representatives of the mononuclear phagocyte and DC system (16, 18, 19). In addition to two subsets of CD207⁺LCs (representing steady-state LCs and neo-appearing CD1c⁺LCs), also CD207^-eDCs, outnumbering steady-state LCs, were described (16, 37). An overview of markers for classifying these DC subsets is shown in Supplementary Table 1. Their lineage relationship and the nature of the epithelial differentiation signals remained poorly defined. Since TGF-β1 promotes CD207⁺LC differentiation from blood monocytes (22) and cDC2s (21, 38) in vitro, we first quantified TGF-β1 and downstream p-SMAD2/3 in non-involved versus lesional epidermal sites. This revealed equivalent expression levels (Supplementary Figures 2A, B). This contrasted with the previously observed aberrant high BMP7 and downstream p-SMAD1/5/8 signal throughout the multilayered lesional epidermis (39, 40). Thus, both TGF-β1 and BMP7 signaling pathways are active in psoriatic epidermal cells. Since CD207^-eDCs in the inflamed epidermis reportedly express Axl mRNA (16) and Axl is similarly expressed by LCs (29), we reasoned that Axl protein might represent a useful pan-DC marker in stratified epithelia/epidermis, potentially enabling to delineate epithelial DC subset differentiation. To address this question, we first studied Axl expression by blood cDC2 in vitro, followed by human tissue analysis, and using a differentiation model of human progenitor cells (see below). Purified CD1c⁺ blood DCs (i.e. termed cDC2) were cultured in the presence of GM-CSF for 5 days, and the effects of addition of BMP7 and/or TGF-β1 on cell differentiation was analyzed using flow cytometry. Initial experiments revealed that BMP7 shows little or no effect on Axl expression by cDC2s, when added to GM-CSF supplemented cultures (Supplementary Figure 3B). Conversely, Axl was induced to high levels by large percentages of blood cDC2s in response to TGF-β1 in vitro. In these differentiation cultures, Axl⁺ cells were sub-divided into two distinct cell subsets based on CD207 expression (Figure 3A). Pre-stimulation of cells with BMP7 for 3 days, followed by TGF-β1 for additional 48 h led to the preferential generation of Axl⁺CD207^-cDC2 (Figures 3A, B). Both Axl⁺CD207^- and Axl⁺CD207⁺ cells generated from blood cDC2s in presence of TGF-β1 plus BMP7 exhibited cDC2 marker characteristics (CD1c⁺CLEC10A⁺CD11c^hiSIRPα⁺CD5^+/-) (Figure 3C). The CD207⁺ cells in TGF-β1 only cultures showed equivalent expression of cDC2-affiliated CD1c, CLEC10A and SIRPα, but less HLA-DR relative to Axl⁺cDC2s (Supplementary Figure 4). These phenotypic data of cDC2-derived CD207⁺ cells are confirmative to previous observations (21). Given the previous description of a minor subset of Axl⁺CD1c⁺ blood DCs (10), and of TGF-β synthesis by cDC2 (41), we next asked whether the constitutive Axl expression by Axl⁺CD1c⁺ blood DCs might similarly be regulated by TGF-β1. TGF-β1-induced Axl in blood cDC2s can be blocked by the addition of an inhibitor of canonical TGF-β1 signaling (ALK4/5/7 inhibitor SB 431542). The low constitutive Axl expression by blood cDC2s observed in most samples analyzed in the absence of TGF-β1 was also consistently reduced by the ALK4/5/7 inhibitor, indicating that canonical TGF-β1 signaling may maintain constitutive Axl expression by blood cDC2s. Given that TGF-β1 signals through BMPR1a/ALK3 to promote LC differentiation (42, 43), we added dorsomorphin, an inhibitor of BMP signaling, targeting type 1 BMP receptors ALK2/3/6 to parallel cultures. Basal and TGF-β1-induced Axl expression remained unaffected by dorsomorphin (Figure 3D). Moreover, since KLF4 interferes with TGF-β1 signaling in monocytes (33, 34), we asked whether short-term stimulation of cDC2 with GM-CSF/IL-4, i.e. factors promoting KLF4^hi moDCs (Figures 2A, C), might abrogate Axl induction. This was indeed the case (Supplementary Figure 6). In addition to strong BMP7 signaling, psoriatic lesional cells are marked by p38MAPK activation (44), and its activation in keratinocytes promotes psoriasis-like lesion formation in mice (45). Recently, the c-Jun/AP-1 signaling complex downstream of p38MAPK has been found to be highly activated in human psoriatic lesions. Particularly, cDC2s recruited to psoriatic inflammatory lesions show strong c-Jun activation, resulting in enhanced pro-inflammatory cytokine production (46). Therefore, we analyzed the role of p38MAPK activation on blood cDC2s undergoing TGF-β1-dependent Axl⁺ cell differentiation. Since p38MAPK signaling is known to be activated in GM-CSF stimulated cells in vitro [reviewed in (47)], we assessed the contribution of p38MAPK signaling in cDC2 differentiation, by adding a small molecule inhibitor. Short-term addition of the p38MAPK inhibitor SB 203580 diminished Axl⁺cDC2 in favour of CD207⁺ LC generation among Axl⁺ cells (Figure 3E). Together, these data revealed that two signals aberrantly induced in the lesional psoriatic epidermis (i.e. BMP7/BMPR1a and p38MAPK) promote Axl⁺cDC2 differentiation from blood cDC2. Moreover, diminishment of these lesional signals promotes CD207⁺ LCs at the expense of Axl⁺cDC2s.

Above experiments revealed that anti-Axl stainings can be used to delineate Axl⁺cDC2 and to study their differentiation into LCs. BMP7 and p38MAPK promoted Axl⁺cDC2 at the expense of LCs. Previous studies showed p38MAPK activation in response to BMPR1A signaling in other cell types (48), opening the possibility that p38MAPK is activated downstream of BMP7 signaling in cDC2 differentiation. We next asked whether above-mentioned in vivo occurring psoriatic CD207^-eDCs (37) exhibit molecular resemblance with activated (TLR2-stimulated) BMP7-driven LC/DC-like cells generated in cultures of CD34⁺ progenitor cells (40). As also shown for neo-appearing lesional epithelial DCs and LC-like cells, these latter cells expressed CD1a, lacked Birbeck granules, and exhibited a CD207^loCD1c⁺CD206⁺ marker profile (40). Data analysis via the Ingenuity Pathway Analysis platform indeed revealed that both cell fractions exhibit pronounced commonalities, as 75% of Canonical Pathways and 65% of Upstream Regulators described on the basis of BMP7-associated genes overlapped with the eDCs found in the psoriatic patients (Figure 4A). Among the common induced pathways, we identified IL-17A and ERK/MAPK signaling, both known to be fundamentally involved in atopic dermatitis and psoriasis (44, 49, 50). Subsequent IPA-based analysis identified TGF-β, BMP7, NOTCH1, p38MAPK/RelB and CDH1/CTNNB1 as common significant Upstream Regulators (Figure 4A and Supplementary Tables 5–8). Therefore, lesional CD207^-eDCs indeed exhibit close resemblance with BMP7 driven CD207^lo LC/cDC2-like cells.

In subsequent analyses we aimed to identify Axl⁺cDC2 within the epidermis of psoriatic cutaneous lesions. GENEVESTIGATOR- based comprehensive analysis of a publicly available transcriptomic dataset comprising healthy, psoriatic and resolved epidermal tissue (51) showed up-regulation of Axl, CD11c and SIRPα in the diseased skin sections, markers expressed by Axl⁺cDC2 (Figure 4B). We next performed quantitative immunofluorescence analysis of healthy and psoriatic human skin biopsies for CD207, CD1c and Axl using triple stainings. Being aware that epidermal keratinocytes of human healthy skin biopsies reportedly express Axl (29) and potentially gain low levels of CD1c under inflammatory conditions, only CD1c^hi cells were considered for the analysis. Only CD207^hi cells were considered as epidermal LCs. Transcriptomic analysis of single-cell RNAseq datasets (Supplementary Figure 7A) and total RNAseq datasets (Supplementary Figure 7B) of healthy skin sections revealed (i) higher Axl expression in DCs compared to epidermal keratinocytes and (ii) lack of CD1c in healthy keratinocytes. Indeed, computerized microscopy and the follow-up quantitative analysis revealed the presence of a spectrum of epidermal DC subsets in psoriatic human skin sections (Figure 4C). In the increased stratum spinosum, we localized two distinct cell types marked as CD207⁺CD1c⁺Axl⁺ (classical steady-state LC morphology) and CD207^low/-CD1c⁺Axl⁺. Within the papillae, we found CD1c⁺Axl⁺CD207^+/- cells, which might translocate from the epidermal to the dermal region. In comparison, in healthy skin biopsies, we only occasionally found Axl⁺CD1c⁺CD207^- cells, and the majority of the star-like shaped CD207^hiCD1c^-Axl⁺ LCs resided in the epidermal stratum spinosum. Next, we performed statistical evaluation by comparative analysis using the densities of marker-positive cells as variables. Healthy (non-involved) and psoriatic skin samples were compared for CD1c, Axl and CD207. The CD1c⁺Axl⁺ double positive cell population was further assessed for co-expression of LC marker CD207. Overall, we found that the psoriatic skin harbours CD1c⁺Axl⁺ cells, the majority thereof lacking or only expressing low levels of LC-marker CD207 (Figure 4D). These observations are consistent with previously reported CD207^- eDCs exhibiting Axl mRNA expression and outnumbering LCs in the lesional psoriatic epidermis (16).

The non-classical NF-kB family member RelB was both implicated in p38MAPK signaling in DCs and in murine cDC2 differentiation. Specifically, RelB is induced downstream of p38MAPK signaling in CD1a⁺DCs/LCs in cultures of human progenitor cells (52), and is essential for myeloid-related DC differentiation in mouse and human [murine splenic CD4⁺ESAM⁺cDC2 (53) and CD11b⁺CD8α^-DC (54); human monocyte-derived interstitial/dermal CD11b⁺DC (55)]. Conversely RelB is dispensable for LC differentiation, and epidermal LCs lack RelB (54, 55). Its positive effects on human interstitial/dermal-type DC differentiation was mediated via the promotion of monocyte intermediates pre-generated in cultures of human CD34⁺ hematopoietic progenitor cells (55). Our in vitro blood cDC2 differentiation cultures revealed that RelB is expressed by a portion of GM-CSF/TGF-β1 stimulated cDC2s in a p38MAPK dependent manner (Figure 5A), with p38MAPK inhibition resulting in diminished CD80 and CD86 expression (Figure 5B). RelB⁺ cells emerging in cultures of blood cDC2 in response to GM-CSF/TGF-β1 lacked both Axl and CD207 (Figures 5C, D). RelB stainings of the inflamed skin have to our knowledge not been reported previously. Immunohistology revealed numerous RelB⁺CD1c⁺ cells in the lesional psoriatic skin, but not in healthy control (Figure 5E), and these cells predominantly occurred in the dermal compartment (Figure 5E). In conclusion, RelB can be induced in blood cDC2s and these cells lack Axl and CD207. RelB⁺CD1c⁺ cells may therefore represent a distinct cDC2 differentiation pathway in the inflamed skin.

To further delineate the lineage relationship and signal requirement of above described epithelial Axl⁺DC/LC subsets, we utilized differentiation cultures of CD34⁺ hematopoietic progenitor/stem cells. cDC2-like cells can be generated in vitro in association with cDC1 and pDCs (56). However, co-generation of cDC2-like cells together with LCs has to our knowledge not been reported. TGF-β1 induces LC differentiation when added to GM-CSF plus FLT3-L and TNFα supplemented cultures of CD34⁺ HPCs. Adding BMP7 instead of TGF-β1 to these cultures also led to LC differentiation (43). These BMP7-LC generation cultures showed higher cellularity and more numerous large homotypic LC-type clusters than TGF-β1-LC generation cultures (43). mRNA profiling and subsequent FACS analysis revealed among others expression of the cDC2 markers CD1c and CLEC10A (12, 57) by CD207⁺ cells from BMP7-LC but not from TGF-β1-LC generation cultures (40). Above pathway analysis confirmed that psoriatic eDCs closely resemble BMP7-LCs (Figure 4A). Conversely LC-like cells generated in TGF-β1-LC cultures phenotypically resembled steady-state LCs (CD1c^-CD207^hi). Interestingly, CD1a⁺ cells lacking CD207 accumulated in BMP7-LC generation cultures, and were much less frequent in TGF-β1-LC generation cultures (Figure 6A) (40, 43). These cells remained unclassified. Flow cytometry analysis revealed that they exhibit all analyzed cDC2 marker characteristics (CLEC10A⁺CD1c⁺SIRPα⁺CD11c⁺HLA-DR⁺CD5^lo; Figure 6B) (11, 12). We previously observed that CD1a⁺CD207⁺ BMP7-LCs exceed TGF-β1-LCs in their capacity to induce regulatory T cell generation from naïve allogeneic CD4⁺ T cells (39). A similar effect was seen when analyzing flow sorted CD1a⁺CD207^- cells [i.e. CD1a⁺CD207^- cells from BMP7-LC exceeding those from TGF-β1-LC cultures in regulatory T cell (Treg) induction (Supplementary Figure 8)]. Therefore, BMP7 in combination with additional cytokines (GM-CSF, FLT3-L and TNFα) co-induces cDC2- and LC-like cell generation from CD34⁺ HPCs, and these cells are potent inducers of Tregs, corroborating an anti-inflammatory effect of BMPR1a in murine CD11c⁺ DCs (39, 42). Considering their regulatory function and the presence of numerous large cell clusters in BMP7 cultures (Figure 6C), morphologically resembling previously described E-cadherin-dependent LC clusters in TGF-β1-LC generation cultures (58), we further asked whether CD1a⁺CD207^-cDC2-like cells express E-cadherin and its intracellular binding partner β-catenin. We found high percentages of E-cadherin among CD207^- cells generated in BMP7 cultures, and these cells were more frequently observed in BMP7- than in TGF-β1-supplemented cultures. Moreover, BMP7 supplemented cultures showed in 6 of 9 independent experiments higher percentages of β-catenin compared to TGF-β-supplemented cells (Figure 6D). We further asked whether the in vitro generated cDC2-like cells may show TGF-β1-inducible Axl expression, as observed for blood cDC2s (Figure 3A). In these experiments (in Figure 6E) we first cultured progenitor cells in the presence of BMP7 (plus basic cytokines GM-CSF, TNFα, FLT3L and SCF), followed by TGF-β1 addition for additional 48 h. Indeed, a portion of day 5 generated cDC2-like cells in the presence of BMP7 acquired Axl in response to short term (48 h) TGF-β1 stimulation. Specifically, we observed the induction of two cell populations, i.e. CD207⁺Axl⁺ and CD207^-Axl⁺ (Figure 6E), both exhibiting cDC2 marker characteristics (CLEC10A⁺CD1c⁺SIRPα⁺CD5^+/-CD11c⁺HLA-DR⁺; Figure 6F). In subsequent experiments, we asked whether CD207^-cDC2-like cells may differentiate into CD207⁺LCs, and whether canonical TGF-β1 signaling might repress cDC2 in favour of LC characteristics. Adding TGF-β1 to BMP7-cDC2 cultures indeed induced a phenotypic shift of CD1a⁺CD207^- to CD1a^hiCD207⁺ cells (Supplementary Figure 9A) and slightly repressed cDC2 markers (CD1c, CLEC10A and SIRPα; Supplementary Figure 9B). Since TGF-β1 can signal through both ALK5 and ALK3/BMPR1a, we performed an inverse experiment in which we inhibited canonical ALK5 signaling in TGF-β1-supplemented LC generation cultures. Adding an ALK5 inhibitor promoted the expression of all three cDC2 markers analyzed (Supplementary Figure 9C), supporting that BMPR1a/ALK3 promotes, whereas canonical TGF-β1-ALK5 represses cDC2 marker characteristics. However, these data do not prove that TGF-β1 signaling inhibits cDC2 markers by CD207⁺CD1a^hi LCs.

Axl is constitutively expressed by a subset of peripheral blood CD1c⁺ cDC2s. Given that above in vitro generated cDC2-like cells co-express E-cadherin and Axl (Figures 6D, E), we asked whether E-cadherin is similarly expressed by Axl⁺ blood cDC2s. We confirmed that purified CD1c⁺ DCs express cDC2 associated markers CLEC10A, CD1c, SIRPα with a subset being Axl⁺ (10, 11) (Figure 7A). E-cadherin can indeed be detected on a subset of blood cDC2s and these cells were enriched among the CD5⁺ cell subset (9) (Figure 7B), and overlapped with Axl⁺ cells (Figure 7C).

Umbilical cord blood was obtained from healthy donors during full-term delivery within the ethical approval (EK 26-520 ex 13/14) obtained from the Medical University of Graz. The isolation of CD34⁺ hematopoietic progenitor cells was performed by magnetic sorting with the EasySep human CD34 positive selection kit (Stem Cell Technologies) according to manufacturer’s instructions. Isolated CD34⁺ cells were pre-expanded in serum-free X-VIVO15 (Lonza) medium, in the presence of 50 ng/mL SCF, 50 ng/mL FLT3-L and 50 ng/ml TPO for 3-4 days. For subsequent LC differentiation, 5x10⁴ pre-expanded CD34⁺ cells were cultured in 24-well tissue culture plates in serum-free CellGro GMP DC medium (CellGenix, Freiburg, Germany) supplemented with 100 ng/mL GM-CSF, 2.5 ng/mL TNFα, 50 ng/mL FLT3-L, 20 ng/mL SCF, and 1 ng/mL TGF-β1 or 200 ng/mL BMP7 for 7 days. On day 3, fresh BMP7 or TGF-β1 was added. Monocytes were isolated using the CD14 MicroBeads (Miltenyi). For CD1c⁺ DC culture, PBMCs of two donors were pooled and cells were isolated using the human CD1c (BDCA-1)⁺ Dendritic Cell Isolation Kit (Miltenyi) according to manufacturer’s protocol. The purity of all isolated fractions was analyzed by flow cytometry. For DC generation, purified cells were stimulated with 100 ng/mL GM-CSF and 35 ng/mL IL-4. Macrophage differentiation was induced by the addition of 100 ng/mL M-CSF and 2 ng/mL IL-6, LCs were generated from purified CD1c⁺ blood DCs in response to 100 ng/mL GM-CSF and 1 ng/mL TGF-β1 or 200 ng/mL BMP7. FACS analysis was performed after 5-6 days. These cell culture experiments were performed in RPMI 1640 media supplemented with 10% FBS, 2.5 mmol/L GlutaMAX (Invitrogen, Grand Island, NY) and 125 U/ml penicillin/streptomycin (PAA, Pasching, Austria). Cytokines and reagents are listed in Supplementary Table 2. Inhibitors for BMP (dorsomorphin), classical TGF-β1 (SB431542), and p38MAK (SB203580) signaling were titrated to secure viability of cultured cells (the chosen concentrations are indicated in Supplementary Figure 5).

The mixed leukocyte reaction assay was conducted as previously described (39). Day 7 generated CD1a⁺CD207^- cells of CD34⁺ cell-derived TGF-β1 and BMP7-LC cultures were enriched by magnetic sorting. 2.7x10⁴ sorted cells were co-cultured with 1x10⁵ purified allogeneic CD4⁺CD45RA⁺ naive T cells in triplicates. The assay was performed using RPMI 1640 media supplemented with 10% FBS, 2.5 mmol/L GlutaMAX (Invitrogen, Grand Island, NY) and 125 U/ml penicillin/streptomycin (PAA, Pasching, Austria). The cells were analyzed on day 5 by flow cytometry.

The viability of the cells was critically assessed by microscopy. Only samples in which the cells looked healthy were used for further analysis. For flow cytometry analysis, cells were stained and analyzed as previously described (39, 40). In brief, cells were harvested and washed with PBS. Prior to antibody staining, Fc receptors were blocked. For intracellular Treg and β-catenin stainings the FoxP3 staining buffer set (ThermoFisher, Waltham, Mass) was used. Intracellular RelB was analyzed using the FIX&PERM kit (Nordic MUbio, Susteren, the Netherlands) according to the manufacturer’s instructions. All samples were recorded using the LSRFortessa (BD Biosciences) and analyzed with the FlowJo software. All antibodies used are listed in Supplementary Table 3.

Paraffin-embedded biopsies were provided by the department of Dermatology, Medical University of Graz. The study was approved by the Ethical Committee of the State of Carinthia, Austria (Dithranol study ClinicalTrials.gov no. NCT02752672; approval number A23/15). Informed consent was provided to all patients in accordance with the Declaration of Helsinki.

For cell staining a polyclonal rabbit antiactivated Notch-1 (Abcam, Cambridge, United Kingdom) antibody, a polyclonal rabbit anti-KLF4 (Sigma-Aldrich, Cat#HPA002926) antibody and a p-SMAD2/3 (Cell Signaling Technology, Cat#8828) antibody were used.

Tissues were stained using pAb anti-BMP7 (LSBio, Cat# LS-B4567-50), pAb anti-pSMAD1/5/8 (Cell Signaling Technology, Cat# 9516), anti-TGF-β1 (Novus, Cat# NBP2-22114), anti-pSMAD3 (Abcam, Cat# ab118825) and anti-RelB (Cell Signaling Technology, Cat# 10544). Goat anti-Rabbit-Cy3 (Jackson ImmunoResearch Labs, Cat# 111-165-144) and horse anti-Mouse-Dy488 (Vector Laboratories, Cat# DI-2488) were used as secondary antibodies. Nuclei were counterstained with DAPI (Molecular Probes). Pictures were taken by the Leica DM4000B microscope and ZEISS LSM700 confocal microscope (Carl Zeiss Microscopy, White Plains, NY) and further processed using Adobe Photoshop CS5.

For analyzing CD1c, Axl and CD207 co-localisation in human skin sections, TissueFAXS platform (TissueGnostics, Vienna, Austria) was used for acquisition and quantitative analysis. Paraffin-embedded 4 µm sections of healthy (n=4) and active psoriatic (n=4) tissue sections were used for stainings. For the detection of CD1c, a monoclonal mouse antibody (Abcam, Cat# ab156708) was used followed by a donkey anti-mouse AF488-conjugated antibody (Jackson ImmunoResearch Labs, Cat# 715-545-150). Axl was detected by a rabbit antibody (Cell Signaling Technology, Cat# 4977), followed by a donkey anti-rabbit Cy3-conjugated antibody (Jackson ImmunoResearch Labs, Cat# 711-165-152). A rat antibody was used to detect CD207 (Eurobio Scientific, Cat#DDX0362B), followed by a donkey anti-rat AF647-labelled secondary antibody (Jackson ImmunoResearch Labs, Cat# 712-605-153). IgG rabbit (Calbiochem, Cat#N101), IgG mouse (Millipore Corporation, Temecula, Ca, USA, #N103) and IgG2a rat (Thermo Fisher Scientific, Cat# 14-4321-81) were used for isotype control. Nuclei counterstaining was performed using DAPI. All tissue slides were mounted using Dako Fluorescence Mounting Medium (Agilent, Inc, Santa Clara, California) and stored at -20°C until acquisition. Whole skin sections were recorded in the DAPI (nuclei), TRITC (Axl), GFP (CD1c) and Cy5 (CD207) channels using the 20x objective by automated tissue FAXS system. Quantitative analysis was performed using the TissueQuest software (version 7.1.1.123) (TissueGnostics, Vienna, Austria). The epidermis was selected as region of interest (ROI) and considered for quantitative analysis. Areas of parakeratosis, staining artefacts and high auto-fluorescence regions were excluded from the analysis. Patient-specific thresholds on scatter plots for detection of marker-positive cells were determined by three independent and experienced observers. Additionally, isotype staining of healthy and psoriatic skin biopsies were used as controls. Using TissueQuest software, for each patient the density values of marker-positive cells were exported and statistically analyzed using GraphPad Prism 5.0.0.286.

GENEVESTIGATOR is a platform which comprises a compendium of publicly available and curated microarray, mRNASeq, and single-cell Seq datasets. The power of GENEVESTIGATOR alone and in combination with tissue image cytometry by TissueFAXS was previously shown by us (74–77). Four different datasets (GSE118165: n_total=157; GSE115736: n_total=42; GSE75042: n_total=9; GSE107011: n_total=127) of the mRNA Seq Gene Level Homo sapiens (Ref: Ensembl 97, GRCh38.12) platform were analyzed. From the multitude of cell types, datasets attributed to cDCs and CD14 monocytes (n=28) were filtered. Expression values (log2) of selected genes of interest were extracted for each sample and used for follow-up statistical analysis. Statistical significance was determined, and graphical visualisation was achieved using GraphPad Prism 5.0.0.286.

Expression values (log2) of human healthy and psoriatic epidermis (GSE103489) (Figure 4B) and healthy human skin [total RNA-seq Gene Level Homo sapiens GSE115898/10x scRNA-Seq Gene Level Homo sapiens (GSE147424)] (Supplementary Figures 7A, B) for the genes of interest were exported from GENEVESTGATOR and analyzed using GraphPad Prism 5.0.0.286.

The IPA platform was used to assign differentially expressed genes to common biologically relevant Canonical Pathways and Upstream Regulators. The analysis was based on genes attributed to DCs, found in the psoriatic epidermis (37) and BMP7-LC, specified above. Gene names and the corresponding fold changes (FC) of the DC profile and differentially expressed genes in BMP7-LCs versus TGF-β1-LCs (FC > 1.5) were used for IPA-based Core analysis. Only statistically significant (p-value < 0.05) Canonical Pathways and Upstream Regulators were exported. The ranking is based on the p-value. For identifying overlaps between two datasets VENNY2.1.0 was used. Graphical visualisation was performed by use of the R package “ggplot”.

Statistical analysis was performed using GraphPad Prism Version 8.3.0 and Version 5.0.0.286. For the comparison of two groups Student’s two-tailed t-test was used, for the analysis of multiple groups a one-way ANOVA was performed. p values < 0.05 were considered as significant. All data are represented as means ± SEM.