Transcriptional Landscape of CUT-Class Homeobox Genes in Blastic Plasmacytoid Dendritic Cell Neoplasm

Homeobox genes encode developmental transcription factors regulating tissue-specific differentiation processes and drive cancerogenesis when deregulated. Dendritic cells (DCs) are myeloid immune cells occurring as two types, either conventional or plasmacytoid DCs. Recently, we showed that the expression of NKL-subclass homeobox gene VENTX is restricted to conventional DCs, regulating developmental genes. Here, we identified and investigated homeobox genes specifically expressed in plasmacytoid DCs (pDCs) and derived blastic plasmacytoid dendritic cell neoplasm (BPDCN). We analyzed gene expression data, performed RQ-PCR, protein analyses by Western blot and immuno-cytology, siRNA-mediated knockdown assays and subsequent RNA-sequencing and live-cell imaging. Screening of public gene expression data revealed restricted activity of the CUT-class homeobox gene CUX2 in pDCs. An extended analysis of this homeobox gene class in myelopoiesis showed that additional CUX2 activity was restricted to myeloid progenitors, while BPDCN patients aberrantly expressed ONECUT2, which remained silent in the complete myeloid compartment. ONECUT2 expressing BPDCN cell line CAL-1 served as a model to investigate its regulation and oncogenic activity. The ONECUT2 locus at 18q21 was duplicated and activated by IRF4, AUTS2 and TNF-signaling and repressed by BMP4-, TGFb- and IL13-signalling. Functional analyses of ONECUT2 revealed the inhibition of pDC differentiation and of CDKN1C and CASP1 expression, while SMAD3 and EPAS1 were activated. EPAS1 in turn enhanced survival under hypoxic conditions which thus may support dendritic tumor cells residing in hypoxic skin lesions. Collectively, we revealed physiological and aberrant activities of CUT-class homeobox genes in myelopoiesis including pDCs and in BPDCN, respectively. Our data may aid in the diagnosis of BPDCN patients and reveal novel therapeutic targets for this fatal malignancy.


Introduction
Transcription factors (TFs) are key players in the developmental control of hematopoiesis, constituting gene regulatory networks which, in turn, orchestrate differentiation processes [1][2][3].Homeobox genes encode TFs sharing the so-called homeodomain which acts as a platform for both DNA and protein interactions [4].This domain belongs to the helix-turn-helix family of TF motifs and consists of 60 amino acid residues, forming three alpha-helices.According to sequence similarities in the encoded homeodomain, homeobox genes fall into eleven classes (e.g., ANTP, CUT and TALE) and several subclasses (e.g., NKL) [5].Homeodomain TFs play basic roles in tissue development and cell differentiation, including hematopoiesis [6][7][8].To evaluate their role in normal and aberrant lymphoid and myeloid immune and blood cell development, we previously identified so-called "TF-codes".These codes describe physiological activities of TALE-class or NKL-subclass homeobox genes, for example, in developing and mature hematopoietic entities, and hence are named "TALE-code" or "NKL-code" [7,9,10].These benchmark gene signatures illuminate our understanding of regulatory processes in the hematopoietic compartment and enable rapid identification of deregulated homeobox genes in corresponding malignancies [7,9,10].Dendritic cells (DCs) are myeloid immune cells which sense and capture bacteria or virus-infected cells and subsequently display specific antigens to T-cells prior to eliciting immune responses [11].According to their function, DCs are present in most tissues and peripheral blood.Hematopoietically, they derive from either specific progenitors or monocytes.Two types of progenitor-derived DCs are distinguished, namely myeloid or conventional DCs (cDCs) and plasmacytoid DCs (pDCs).They differ in expression of specific TFs, surface markers and cytokine production [12][13][14].
Recently, we identified the physiological expression of the NKL-subclass homeobox gene VENTX exclusively in cDCs and the aberrant expression in acute myeloid leukemia (AML) [15].Here, we report on two CUT-class homeobox genes in pDCs: CUX2 physiologically expressed only in pDCs and ONECUT2 aberrantly activated in blastic plasmacytoid dendritic cell neoplasm (BPDCN).CUT-class members share a specific homeodomain and one-to-three CUT domains.CUT domains comprise about 180 amino acid residues and create five alpha-helices.They also perform DNA and protein interactions, thus extending the regulatory potential of these TFs [16,17].The CUT-class of homeobox genes is divided into three groups, CUX, ONECUT and SATB, and the human genome contains seven CUT-class genes, namely CUX1, CUX2, ONECUT1, ONECUT2, ONECUT3, SATB1 and SATB2 [4].ONECUT2 is normally expressed in the liver, pancreas and brain [18,19].The aberrant expression of ONECUT2 has been reported in prostate and lung cancer, demonstrating the oncogenic potential of this factor when deregulated [20][21][22].
BPDCN is a rare hematopoietic malignancy characterized by clonal proliferation of immature pDCs.Diagnostically, BPDCN has to be distinguished from AML [23].The clinical manifestations of BPDCN are predominant cutaneous involvement with subsequent or simultaneous extension to the bone marrow and peripheral blood while normal pDCs are absent from the skin.The clinical course of BPDCN patients is almost invariably aggressive and fatal [24,25].At present, there is no consensus on the optimal treatment of this disease, highlighting the urgency of identifying novel therapeutic options.Our data may improve the diagnosis and therapy of BPDCN.

Homeobox Gene Expression in cDCs and pDCs
Homeobox genes underpin normal differentiation processes and drive cancerogenesis when deregulated.While investigating NKL homeobox genes in DC development, we recently reported the expression of VENTX exclusive to progenitor-derived cDCs [15].To extend the search for homeodomain TFs involved in DC differentiation, we here screened for homeobox genes differentially expressed in mature DCs.To this end, we compared gene expression profiling data for cDCs and pDCs using public dataset GSE24759 and the associated online tool GEOR, thus revealing 805 genes showing statistically significant differences.These genes included four homeobox genes: CUX2 and ZHX2, specifically expressed in pDCs, and HOXA9 and VENTX activated in cDCs (Table S1, Figure 1).Interestingly, these homeobox genes fall into different groups: CUX2 is a CUT-class member, ZHX2 belongs to the ZHX-family, HOXA9 is part of the HOXA-cluster, and VENTX is a member of the NKL-subclass.
This observation prompted transcriptional analysis of all homeobox gene members in normal immune cells for each of these four groups using RNA-seq data from the Human Protein Atlas and additionally in BPDCN and AML patients using the expression profiling dataset GSE89565 (Figure S1).This revealed aberrantly expressed, closely related homeobox genes, notably the CUT-class homeobox gene ONECUT2, which, though silent in all mature immune cells, was aberrantly activated in BPDCN but not AML patients.Moreover, CUX2 activity was restricted to pDCs, underlining the role of CUT-class homeobox genes in normal and malignant pDCs.
To focus on CUT-class homeobox genes in hematopoietic development, we performed gene expression analyses for all seven CUT-class members in myelopoiesis and identified genes active in myeloid progenitors and mature immune cells (Figure S2).We used public datasets and applied cutoffs reported in our previous studies [10].The gene signature thus generated was termed "myeloid CUT-code" and is shown in Figure 2.This code demonstrated that CUX1 and SATB1 were expressed in all entities analyzed.CUX2 was expressed in pDCs and additionally in lymphoid and myeloid primed progenitors (LMPPs), while ONECUT1, ONECUT2 and SATB2 remained silent in the complete myeloid compartment.ONECUT3 activity was restricted to mature myeloid cells including granulocytes, monocytes and megakaryocytes.Of note, there is an ongoing discussion about the course of DC development [26].In mice, cDCs derive from DC progenitors, while pDCs may derive from CLPs, thus representing lymphoid cells [27].Our presented system for human myelopoiesis depicts the classical model which may require correction when this puzzle is solved [13].Taken together, our survey of homeobox genes in DCs and myelopoiesis revealed a specific role of CUT-class homeobox genes CUX2 and ONECUT2 in pDCs and the derived malignancy BPDCN, respectively.In the following, we focused on these two genes in BPDCN.This observation prompted transcriptional analysis of all homeobox gene members in normal immune cells for each of these four groups using RNA-seq data from the Human Protein Atlas and additionally in BPDCN and AML patients using the expression profiling dataset GSE89565 (Figure S1).This revealed aberrantly expressed, closely related homeobox genes, notably the CUT-class homeobox gene ONECUT2, which, though silent in all mature immune cells, was aberrantly activated in BPDCN but not AML patients.Moreover, CUX2 activity was restricted to pDCs, underlining the role of CUT-class homeobox genes in normal and malignant pDCs.
To focus on CUT-class homeobox genes in hematopoietic development, we performed gene expression analyses for all seven CUT-class members in myelopoiesis and identified genes active in myeloid progenitors and mature immune cells (Figure S2).We used public datasets and applied cutoffs reported in our previous studies [10].The gene Comparative gene expression profiling analysis of primary pDC and cDC samples using public dataset GSE24759 and online tool GEOR.Genes showing significant differences in their expression level (N = 805) are indicated in blue and red in the volcano plot (above).Homeobox genes CUX2, HOXA9, VENTX and ZHX2 exhibited differential expression (below).
Since LL-100 excludes BPDCN models, we adopted the BPDCN cell line CAL-1 established in 2005 from a 70-year-old male diagnosed with blastic NK-cell lymphoma which is now called BPDCN [28,29].A combined analysis of public gene expression profiling datasets containing CAL-1 (GSE112209) and 20 AML control cell lines (GSE59808) indicated CUX2 and ONECUT2 activity in CAL-1 (Table S2).Furthermore, these data showed raised expressions of TCL1, TCF4, CD303/CLEC4C, CD303/NRP1 and NCAM1 and downregulation of MPO, CD3 and CD34 for CAL-1, corresponding to reported activities of these established BPDCN biomarkers [23,30].In addition, we performed RQ-PCR analysis of selected cell lines, including CAL-1 (pDC/BPDCN), MUTZ-3 (cDC/AML), and EOL-1, KG-1 and THP-1 (AML), and of primary cell samples including hematopoietic stem cells, monocytes and DCs (Figure 3).Accordingly, CUX2 was expressed in CAL-1 and primary DCs, while ONE-CUT2 expression was restricted to CAL-1.Western blot and immuno-cytology analyses confirmed CUX2 and ONECUT2 expression in CAL-1 at the protein level (Figure 3).Taken together, cell line CAL-1 expressed both CUT-class homeobox genes CUX2 and ONECUT2 in addition to basic BPDCN biomarkers, endorsing its use as a model to investigate the potential oncogenic role of ONECUT2 in BPDCN.

Deregulation of ONECUT2 in BPDCN Cell Line CAL-1
In hematopoietic malignancies, chromosomal and genomic rearrangements frequently underlie aberrant oncogene activation [31,32].The ONECUT2 locus is located at chromosomal band 18q21.However, karyotypes reported for BPDCN patients and cell line CAL-1 discount recurrent rearrangements at that locus [28,33,34].To identify genomic copy number alterations in CAL-1, we performed genomic profiling (Figure S4).These data revealed genomic duplication of the ONECUT2 locus in CAL-1 cells (Figure 4A), plausibly supporting its aberrant expression.
To identify TFs involved in the regulation of ONECUT2 in CAL-1, we screened data from the UCSC genome browser showing potential TF binding sites at ONECUT2 (Figure 4B).The results indicated the potential regulatory impact of IRF factors via three binding sites.Our comparative cell line gene expression profiling data showed elevated expression of pDC-associated IRF8 and IRF4 in CAL-1 (Table S2), which was confirmed by RQ-PCR analysis (Figure 4C).SiRNA-mediated knockdown experiments discounted such a role for IRF8 while demonstrating an activating impact of IRF4 for ONECUT2 expression (Figure 4C).Moreover, our genomic profiling data showed genomic gains including IRF4 at 6p25 and IRF8 at 16q24 plausibly underlying their elevated expression (Figure 4D).Taken together, the aberrant expression of ONECUT2 in BPDCN cell line CAL-1 is driven by focal genomic aberrations, respectively, targeting ONECUT2 and IRF4-the latter operating as a transcriptional activator.
To identify additional factors regulating ONECUT2 expression in BPDCN, we analyzed the gene expression profiling data of CAL-1 in comparison to AML control cell lines in more detail (Table S2).By scrutinizing activity of the top-1000 differentially overexpressed and downregulated genes, we observed raised levels of SMAD9, TNF, TNFRSF11A, CHRDL1 and AUTS2 and reduced expression of BMP2K, SMAD1 and FGFR1.These findings indicated potential impacts of BMP/TGFb-, FGF-and TNF-signaling and of chromatinregulator AUTS2.
We tested the impacts of these factors by treatment of CAL-1 cells with BMP4 and TGFb which resulted in slightly reduced ONECUT2 expression, while siRNA-mediated knockdown of the BMP-signaling inhibitor SMAD9 reduced ONECUT2 activity (Figure 5A).FGF2-treatment showed no effect, while FGF9 inhibited ONECUT2 expression significantly (Figure 5A).Treatment of CAL-1 with TNF activated ONECUT2 together with the TNF-control gene TNFAIP3.However, siRNA-mediated knockdown of overexpressed TNFRSF11A failed to impact ONECUT2 expression (Figure 5B).Finally, siRNA-mediated knockdown of AUTS2 inhibited expression of ONECUT2, indicating an activating role (Figure 5C).Interestingly, our genomic profiling data showed an intragenic deletion of AUTS2 which may support its expression and/or activity (Figure 5D).Moreover, ONECUT2-positive BPDCN patients expressed reduced levels of chromatin-repressor PCGF5, which reportedly competes with chromatin-activator AUTS2 (Figure 5C, Table S3) [35].Taken together, our results indicated that TNF-signaling activated while BMP4-, FGF9-and TGFb-signaling inhibited ONECUT2 expression in BPDCN cell line CAL-1.Altered activities of these pathways thus may contribute to aberrant ONECUT2 expression.Furthermore, AUTS2 and PCGF5 may (de)regulate ONECUT2 activity via chromatin modifications in the tumor cells.

Oncogenic Activities of ONECUT2 in BPDCN
In prostate cancer, ONECUT2 operates as an oncogene by activating expression of SMAD3 which regulates tumor growth under hypoxic conditions [21].Here, siRNAmediated knockdown experiments in CAL-1 demonstrated that ONECUT2 activated SMAD3 in BPDCN as well (Figure 6A).Furthermore, our comparative gene expression profiling data indicated reduced expression of CDKN1C and CASP1 in CAL-1 (Table S2).CDKN1C inhibits proliferation and CASP1 mediates pyroptosis-a form of cell death triggered by proinflammatory signaling, thus representing tumor suppressor genes [36,37].Knockdown of ONECUT2 resulted in elevated expression of CDKN1C and CASP1, demonstrating that ONECUT2 inhibits both genes (Figure 6A).

Oncogenic Activities of ONECUT2 in BPDCN
In prostate cancer, ONECUT2 operates as an oncogene by activating expression of SMAD3 which regulates tumor growth under hypoxic conditions [21].Here, siRNA-mediated knockdown experiments in CAL-1 demonstrated that ONECUT2 activated SMAD3 in BPDCN as well (Figure 6A).Furthermore, our comparative gene expression profiling data indicated reduced expression of CDKN1C and CASP1 in CAL-1 (Table S2).CDKN1C inhibits proliferation and CASP1 mediates pyroptosis-a form of cell death triggered by proinflammatory signaling, thus representing tumor suppressor genes [36,37].Knockdown of ONE-CUT2 resulted in elevated expression of CDKN1C and CASP1, demonstrating that ONECUT2 inhibits both genes (Figure 6A).To analyze ONECUT2 target genes in BPDCN systematically, we performed RNA sequencing of CAL-1 cells treated for ONECUT2 knockdown in triplicate.The most significantly altered genes are listed in Table 1.Here, we focused on EPAS1 which encodes a factor orchestrating the cellular physiology under hypoxic conditions [38].RQ-PCR and Western blot analyses showed elevated EPAS1 levels in CAL-1 and siRNA-mediated knockdown experiments confirmed the activating impact of ONECUT2 on EPAS1 (Figure 6B).Furthermore, expression of EPAS1 and its cofactors ARNT and ARNTL correlated with ONECUT2 in BPDCN patients and cell line CAL-1, respectively, supporting functional relevance of these findings (Tables S2 and S3).Interestingly, LL-100 RNA-seq data of leukemia/lymphoma cell lines demonstrated elevated EPAS1 expression in ONECUT2-positive anaplastic large cell lymphoma (ALCL)-derived cell lines (Figure 6D).In contrast to BPDCN cell line CAL-1, ALCL cell line DEL showed a genomic gain at 2p21 hosting the EPAS1 locus (Figure 6E).Thus, aberrantly enhanced EPAS1 expression is driven by genomic rearrangement in ALCL subsets and by ONECUT2, likely affecting both BPDCN and ALCL (Figure S3).Moreover, EPAS1 overexpression correlated with the presence of hallmark fusion gene NPM1::ALK in ALCL patients and cell lines (Figure 6F).Taken together, we identified specific target genes of ONECUT2 in BPDCN, including repressed CDKN1C and CASP1, and activated SMAD3 and EPAS1.SMAD3 reportedly supports the growth of prostate cancer cells under hypoxic conditions [21], further supporting the notion that hypoxic adaptation may represent an important oncogenic function of ONECUT2 and its target genes SMAD3 and EPAS1 in BPDCN.
Next, we analyzed the function of ONECUT2 in BPDCN cell line CAL-1 by live-cell imaging.SiRNA-mediated knockdown of ONECUT2 reduced the proliferation but spared survival (Figure 7A).Thus, ONECUT2 supports proliferation, correlating with repression of cell cycle inhibitor CDKN1C.
Treatment of pDC-derived cell line CAL-1 with GM-CSF (alias CSF2) and IL3 overcomes its pathological differentiation arrest, continuing the process of DC maturation [28].Undifferentiated CAL-1 cells aggregate in culture while treatment with GM-CSF and IL3 suppressed this behavior, thus serving as a surrogate to quantify DC differentiation (Figure 7B).The used software tool analyzing live-cell imaging data is unable to identify single cells located in clusters, resulting in reduced cell counts.GM-CSF/IL3 stimulation dispersed the cell aggregates, resulting in increased cell counts.Accordingly, siRNA-mediated knockdown of ONECUT2 enhanced cell aggregation inhibited by GM-CSF/IL3 stimulation (Figure 7B), indicating that ONECUT2 repressed DC differentiation processes.
Finally, to investigate the potential role of the ONECUT2 target gene EPAS1 in cell survival under hypoxic conditions, we reduced the normal oxygen level (20%) to 5%.EPAS1 expression levels were reduced in CAL-1 cells by siRNA-mediated knockdown.Subsequent quantification of apoptotic cells by live-cell imaging demonstrated that EPAS1 supported survival of BPDCN cells growing with reduced oxygen (Figure 7C).This function may play an oncogenic role for tumor cells residing in hypoxic skin lesions.

Discussion
In this study, we investigated CUT-class homeobox genes in myelopoiesis including DCs.CUX2 was physiologically expressed in pDCs and silent in cDCs, while ONECUT2 was aberrantly expressed in BPDCN.Using BPDCN cell line CAL-1 as a model, we characterized the regulatory architecture of ONECUT2 by identifying its upstream factors and pathways and downstream target genes and functions.These results are summarized in Figure 8, showing a pathogenic gene regulatory network around ONECUT2 which drives oncogenic activities in this fatal malignancy.
survival under hypoxic conditions, we reduced the normal oxygen level (20%) to 5%.EPAS1 expression levels were reduced in CAL-1 cells by siRNA-mediated knockdown.Subsequent quantification of apoptotic cells by live-cell imaging demonstrated that EPAS1 supported survival of BPDCN cells growing with reduced oxygen (Figure 7C).This function may play an oncogenic role for tumor cells residing in hypoxic skin lesions.

Discussion
In this study, we investigated CUT-class homeobox genes in myelopoiesis including DCs.CUX2 was physiologically expressed in pDCs and silent in cDCs, while ONECUT2 was aberrantly expressed in BPDCN.Using BPDCN cell line CAL-1 as a model, we characterized the regulatory architecture of ONECUT2 by identifying its upstream factors and pathways and downstream target genes and functions.These results are summarized in Figure 8, showing a pathogenic gene regulatory network around ONECUT2 which drives oncogenic activities in this fatal malignancy.The founding member of CUT-class homeobox genes was identified in Drosophila, where it regulates both cell-type specification and morphogenesis [39,40].Subsequent analyses of CUT-class homeobox genes in vertebrates and humans confirmed similar functions in several tissues, including the hematopoietic system.For example, CUX1 affects both myeloid and lymphoid differentiation, and CUX2 controls HOX gene expression during forelimb development [41,42].ONECUT1 and ONECUT2 regulate early retinal cell fates, while the latter also regulates intestinal cell differentiation [43,44].SATB1 supports selfrenewal of HSCs inhibiting their differentiation and orchestrates erythroid development via regulation of globin genes [45,46].Furthermore, its CUT domain mediates interaction with the matrix attachment region, thus organizing intranuclear structures [47].Collectively, CUT-class homeobox genes are tightly involved in regulation of development and cell differentiation, including immune and blood cells.
In previous studies, we have proposed several types of TF-code, detailing gene signatures for selected TF groups in the hematopoietic compartment.Examples include the NKL-and TALE-codes, which cover NKL-subclass and TALE-class homeobox gene activities for each progenitor and mature immune cell entity [7,9,10].In line with these previously published codes, we here describe the myeloid CUT-code, covering CUT-class homeobox gene activities in developing and mature myeloid cell types.Hence, CUX1 was detected in all entities which may correspond to its reported general impact on myeloid differentiation [42].CUX2 expression was restricted to LMPPs and pDCs, indicating that this CUT-class homeobox gene may specifically regulate the differentiation of pDCs and cDCs.ONECUT2 was physiologically silent in the complete myeloid compartment, demonstrating ectopic activation in BPDCN.In murine hematopoiesis, DCs derive from CDPs or CLPs [13,26,27].However, the NKL homeobox gene VENTX is specifically expressed in human hematopoiesis including cDCs but lacks in the genome of mice and rats [15,53].Thus, gene activities differ in human and murine hematopoiesis, reminding us not to transfer mouse data to humans without analysis of the corresponding human cell types.
In addition to genomic aberrations, we identified four altered pathways which contributed to ONECUT2 expression, including BMP4-, FGF9-, TGFb-and TNF-signaling.BMP-signaling inhibits ONECUT2 in trigeminal sensory neuron differentiation [54], demonstrating that this pathway plays a physiological role in ONECUT2 regulation in some cell types.We failed to detect any regulatory impact for TNFRSF11A (also named RANK); although, this receptor has been shown to activate ONECUT2 in enterocytes of the small intestine [43].In T-cells, ONECUT2 is regulated by TBX21/T-bet [55].However, this factor is not expressed in CAL-1 (Table S2), probably discounting any similar role in BPDCN.In contrast, we identified elevated IRF4 as a transcriptional activator of ONECUT2 in CAL-1 where both IRF4 and IRF8 are highly expressed and targeted by genomic duplication.These genes play essential roles in development and differentiation of cDCs and pDCs, suggesting aberrant consequences when overexpressed [13,56,57].
In breast cancer, ONECUT2 inhibits CASP3 and activates BCL2, which together support survival of these tumor cells [58].This observation corresponds to our results, demonstrating that ONECUT2 inhibits CASP1 in BPDCN; although, ONECUT2 downregulation showed no significant effect on apoptosis.However, pDCs mediate their immune response by production of inhibitory chemokines which contribute to their pyroptotic cell death as observed during inflammation [59].Regulators and actors in pyroptosis include CASP1, ELANE, GSDMD, IRF8 and SMAD3.These genes showed altered expression levels in CAL-1 (Table S2), collectively indicating suppression of pyroptotic cell death in BPDCN.Therefore, their deregulation may contribute to aberrant survival of BPDCN cells.
DCs are mobile immune cells present in most tissues.Thus, they are confronted with various environments with a diversity of oxygen levels.Reportedly, hypoxia induces cell death of monocyte-derived DCs [60].Furthermore, the hypoxic environment reprograms DCs, alters their metabolism and inhibits differentiation and maturation of pDCs, blunting defensive responses to tumor cells [61,62].These data indicate that normal DCs show substantial responses to hypoxia.Our results indicate that malignant pDCs acquired aberrant activities to escape hypoxia-induced cell death.ONECUT2 activates SMAD3 and EPAS1 which mediate survival of tumor cells under hypoxic conditions as shown recently for SMAD3 in prostate cancer and in this study for EPAS1 in BPDCN [21].Furthermore, CAL-1 cells expressed high levels of ARNTL, while ONECUT2-high BPDCN patients expressed elevated ARNT.EPAS1, ARNT and ARNTL are members of the bHLH/PASfamily and are able to interact and cooperate, suggesting functional relevance for their coexpression [38].Moreover, in endochondral ossification-where cartilage scaffolds are replaced by forming bone-EPAS1 activates expression of RUNX2 [63].In BPDCN, the super-enhancer from RUNX2 is frequently translocated to MYC, thus acting as an oncogene-driver in this malignancy [33,34].Therefore, aberrant EPAS1 activity may plausibly enhance MYC expression via the translocated enhancer of RUNX2 in these cells.
Finally, we detected elevated EPAS1 expression in ALK-positive ALCL patients and cell lines.ALCL and BPDCN tumor cells are mostly located in (hypoxic) skin lesions where EPAS1 may support their survival.The availability and application of proven EPAS1inhibitors, like Belzutifan, for treatment of specific cancer types may promote the usage of these drugs for BPDCN and ALCL as well [64].A recently formed consortium aims to reveal novel therapies for BPDCN, including CD123/IL3R-directed drug targeting using tagraxofusp [65].Generally, targeted treatments promise reduced side effects as compared to conventional chemotherapies.However, the pros and cons of novel therapies including Belzutifan have to be validated carefully.Taken together, our data indicate that CUT-class homeobox genes CUX2 and ONECUT2 may be useful biomarkers while hypoxia factors EPAS1 and ARNT/ARNTL may serve as novel therapeutic targets for BPDCN (and ALCL).Accordingly, our findings may illuminate diagnosis of BPDCN and uncover actionable therapeutic targets in this intractable malignancy.

Bioinformatic Analyses of Expression Profiling and RNA-seq Data
Expression data for normal cell types were obtained from Gene Expression Omnibus (GEO, www.ncbi.nlm.nih.gov,accessed on 1 November 2023).We analyzed pDC and cDC samples using dataset GSE24759 and the associated online tool GEO2R for their comparison, revealing significant differentially expressed genes [66,67].Furthermore, we used expression profiling datasets GSE42619, GSE22552, GSE109348, GSE40831 and GSE14879, in addition to RNA-seq dataset GSE69239 and RNA-seq data from The Human Protein Atlas (www.proteinatlas.org,accessed on 1 November 2023) [68].Gene expression profiling data from BPDCN and AML patients were examined using dataset GSE89565 [69].For analysis of cell lines, we exploited RNA-seq data from 100 leukemia/lymphoma cell lines (termed LL-100), available at ArrayExpress (www.ebi.ac.uk/arrayexpress) via E-MTAB-7721.Gene expression values are given as DESeq2 normalized count data [70].ONECUT2 knockdown RNA-seq data are stored at BioStudies (www.ebi.ac.uk/biostudies/) and available via S-BSST1311.

Cell Lines and Treatments
Cell lines are held at the DSMZ (Braunschweig, Germany) and cultivated as described (www.DSMZ.de,accessed on 1 November 2023).All cell lines had been authenticated and tested negative for mycoplasma infection.Modification of gene expression levels was performed using gene specific siRNA oligonucleotides with reference to AllStars negative Control siRNA (siCTR) obtained from Qiagen (Hilden, Germany).SiRNAs (100 pmol) were transfected into 1 × 10 6 cells by electroporation using the EPI-2500 impulse generator (Fischer, Heidelberg, Germany) at 350 V for 10 ms.Electroporated cells were harvested after 20 h cultivation.Cell treatments were performed using 20 ng/mL recombinant BMP4, TGFbeta, TNF, FGF2 and FGF9 for 20 h and 20 ng/mL GMCSF and IL3 for 40 h (R & D Systems, Wiesbaden, Germany).RNA-sequencing analysis was performed in triplicate of CAL-1 cells treated for 20 h by siRNA-mediated knockdown (Eurofins MWG, Konstanz, Germany) using RNAeasy for RNA extraction (Qiagen) and Bioanalyzer for quality control (Agilent Technologies, Santa Clara, CA, USA).For functional testing, treated cells were analyzed with the IncuCyte S3 Live-Cell Imaging Analysis System (Sartorius, Göttingen, Germany).For detection of apoptotic cells, we additionally used the IncuCyte Caspase-3/7 Green Apoptosis Assay diluted at 1:2000 (Sartorius).The cells were incubated in normoxic (20% oxygen, 5% carbon dioxide) or hypoxic (5% oxygen, 5% carbon dioxide) conditions.Live-cell imaging experiments were performed twice with fourfold parallel tests.

Polymerase Chain Reaction (PCR) Analyses
Total RNA was extracted from cultivated cell lines using TRIzol reagent (Thermo Fisher Scientific, Darmstadt, Germany).Primary human total RNA derived from HSCs, monocytes and DCs was purchased from Biochain/BioCat (Heidelberg, Germany).cDNA was synthesized using 1 µg RNA, random priming and Superscript II (Thermo Fisher Scientific).Real-time quantitative (RQ)-PCR analysis was performed using the 7500 Real-time System and commercial buffer and primer sets (Thermo Fisher Scientific).For normalization of expression levels, we quantified the transcripts of TATA box binding protein (TBP).Quantitative analyses were performed as biological replicates and measured in triplicate.Standard deviations are presented in the figures as error bars.Statistical significance was assessed by Student's t-test (two-tailed) and the calculated p-values indicated by asterisks (* p < 0.05, ** p < 0.01, *** p < 0.001, n.s.not significant).

Genomic Profiling Analysis
For genomic profiling, genomic DNA of cell lines was prepared by the Qiagen Gentra Puregene Kit (Qiagen).Labelling, hybridization and scanning of Cytoscan HD arrays were performed by the Genome Analytics Facility located at the Helmholtz Centre for Infection Research (Braunschweig, Germany), using the manufacturer's protocols (Affymetrix, High Wycombe, UK).Data were interpreted using the Chromosome Analysis Suite software version 3.1.0.15 (Affymetrix, High Wycombe, UK) and copy number alterations determined accordingly.

Conclusions
We detected normal and aberrant activities of CUT-class homeobox genes CUX2 and ONECUT2 in pDCs and BPDCN, respectively.ONECUT2 is located in a pathogenic network which contains novel players in this disease.Thus, our results may contribute to improving diagnostic and/or therapy of this fatal malignancy.

20 Figure 1 .
Figure 1.Expression of homeobox genes in DCs.Comparative gene expression profiling analysis of primary pDC and cDC samples using public dataset GSE24759 and online tool GEOR.Genes showing significant differences in their expression level (N = 805) are indicated in blue and red in the volcano plot (above).Homeobox genes CUX2, HOXA9, VENTX and ZHX2 exhibited differential expression (below).

Figure 1 .
Figure 1.Expression of homeobox genes in DCs.Comparative gene expression profiling analysis of primary pDC and cDC samples using public dataset GSE24759 and online tool GEOR.Genes showing significant differences in their expression level (N = 805) are indicated in blue and red in the volcano plot (above).Homeobox genes CUX2, HOXA9, VENTX and ZHX2 exhibited differential expression (below).

Figure 3 .
Figure 3. Expression of CUX2 and ONECUT2 in hematopoiesis and BPDCN.(A) RNA-seq gene expression data for mature immune cells from the Human Protein Atlas shows CUX2 expression exclusively in myeloid or plasmacytoid DCs (above).Gene expression profiling data for BPDCN and AML patients from dataset GSE shows CUX2 expression exclusively in BPDCN patients (middle).RQ-PCR and Western blot analyses show CUX2 expression in pDC cell line CAL-1 and primary

Figure 3 .
Figure 3. Expression of CUX2 and ONECUT2 in hematopoiesis and BPDCN.(A) RNA-seq gene expression data for mature immune cells from the Human Protein Atlas shows CUX2 expression exclusively in myeloid or plasmacytoid DCs (above).Gene expression profiling data for BPDCN and AML patients from dataset GSE shows CUX2 expression exclusively in BPDCN patients (middle).RQ-PCR and Western blot analyses show CUX2 expression in pDC cell line CAL-1 and primary DCs (below).(B) RNA-seq gene expression data for mature immune cells from the Human Protein Atlas discounts ONECUT2 expression in these cells (above).Gene expression profiling data for BPDCN and AML patients from dataset GSE shows ONECUT2 expression exclusively in BPDCN patients (middle).RQ-PCR and immuno-cytology analyses (magnification 400×) show ONECUT2 expression exclusively in pDC cell line CAL-1 (below).

Figure 4 .
Figure 4. Genomic activation of ONECUT2 in BPDCN.(A) Genomic profiling analysis of CAL-1 indicates duplication of 18q21-q23 targeting the ONECUT2 locus at 18q21.31.(B) Analysis of TF binding sites at ONECUT2 using the UCSC genome browser revealed potential impacts of IRF factors.Potential IRF-binding sites are indicated by red arrow heads.(C) RQ-PCR analysis of IRF8 and IRF4 in selected cell lines showed elevated activities exclusive to CAL-1.SiRNA-mediated knockdown experiments revealed an activatory role of IRF4, while IRF8 failed to regulate ONECUT2 expression.p-values are indicated by asterisks (*** p < 0.001, n.s.not significant).Western blot analysis confirmed IRF4 expression at the protein level in CAL-1 cells and its knockdown after siRNA treatment (below).(D) Genomic profiling analysis of CAL-1 indicated copy number gains for IRF4 at 6p25 and IRF8 at 16q24, probably underlying their overexpression.

Figure 5 .
Figure 5. Regulation of ONECUT2 in BPDCN by signaling pathways and chromatin factors.(A) RQ-PCR analyses of ONECUT2 in CAL-1 cells after treatment with BMP4 and TGFb (left), siRNA-mediated knockdown of SMAD9 (middle) and treatment with FGF2 and FGF9 (right).(B) RQ-PCR analyses of ONECUT2 and TNFAIP3 in CAL-1 cells after treatment with TNF (left and middle) and siRNA-mediated knockdown of TNFRSF11A (right).(C) RQ-PCR analyses of ONECUT2 in CAL-1 cells after knockdown of AUTS2 (left).Comparative gene expression profiling data of BPDCN patients expressing high and low/controls ONECUT2 levels for the genes ONECUT2 (middle) and PCGF5 (right).(D) Genomic profiling data for chromosome 7 from BPDCN cell line CAL-1, showing a micro-deletion at the AUTS2 locus (enlargement inserted).p-values are indicated by asterisks (* p < 0.05, ** p < 0.01, *** p < 0.001, n.s.not significant).

Figure 5 .
Figure 5. Regulation of ONECUT2 in BPDCN by signaling pathways and chromatin factors.(A) RQ-PCR analyses of ONECUT2 in CAL-1 cells after treatment with BMP4 and TGFb (left), siRNA-mediated knockdown of SMAD9 (middle) and treatment with FGF2 and FGF9 (right).(B) RQ-PCR analyses of ONECUT2 and TNFAIP3 in CAL-1 cells after treatment with TNF (left and middle) and siRNA-mediated knockdown of TNFRSF11A (right).(C) RQ-PCR analyses of ONECUT2 in CAL-1 cells after knockdown of AUTS2 (left).Comparative gene expression profiling data of BPDCN patients expressing high and low/controls ONECUT2 levels for the genes ONECUT2 (middle) and PCGF5 (right).(D) Genomic profiling data for chromosome 7 from BPDCN cell line CAL-1, showing a micro-deletion at the AUTS2 locus (enlargement inserted).p-values are indicated by asterisks (* p < 0.05, ** p < 0.01, *** p < 0.001, n.s.not significant).

Figure 7 .
Figure 7. Functional analyses of ONECUT2 in BPDCN cell line CAL-1.(A) Analysis of proliferation (left) and apoptosis (right) in CAL-1 cells treated for siRNA-mediated knockdown of ONECUT2 by live-cell imaging.Statistical analysis of the last time point of recording using two-tailed t-test revealed significant differences for proliferation (p = 0.0433) but not for apoptosis.(B) Treatment of CAL-1 cells with GM-CSF and IL3 induced DC differentiation detectable by reduced aggregation of

Figure 7 .
Figure 7. Functional analyses of ONECUT2 in BPDCN cell line CAL-1.(A) Analysis of proliferation (left) and apoptosis (right) in CAL-1 cells treated for siRNA-mediated knockdown of ONECUT2 by livecell imaging.Statistical analysis of the last time point of recording using two-tailed t-test revealed significant differences for proliferation (p = 0.0433) but not for apoptosis.(B) Treatment of CAL-1 cells with GM-CSF and IL3 induced DC differentiation detectable by reduced aggregation of the cells as shown by live-cell images after 40 h (left, magnification 400×).Digital analysis of live-cell images allowed quantification of cell clumping over time: clumped cells were not discriminated leading to decreased cell counts.This approach demonstrated GM-CSF/IL3-induced DC differentiation of CAL-1 which was significantly reduced after ONECUT2 knockdown (two-tailed t-test of last time point of recording: p = 0.0185).(C) RQ-PCR analysis of EPAS1 after siRNA-mediated knockdown of EPAS1 (left).p-value is indicated by an asterisk (*** p < 0.001).Quantification of apoptotic cells by live-cell imaging analysis of CAL-1 treated for EPAS1 knockdown and cultivated under hypoxic conditions.Statistical analysis of the area under curve revealed a significant increase of apoptotic cells after EPAS1 knockdown (p = 0.0016).All live-cell imaging experiments were repeated twice, generating similar results.

Figure 8 .
Figure 8. Gene regulatory network around aberrantly expressed ONECUT2 in BPDCN, summarizing the results of this study.Regulatory relationships are indicated by lines.Upstream regulators

Figure 8 .
Figure 8. Gene regulatory network around aberrantly expressed ONECUT2 in BPDCN, summarizing the results of this study.Regulatory relationships are indicated by lines.Upstream regulators including signaling pathways, genomic aberrations, TFs and chromatin factors are shown.Downstream functions include activation of proliferation and hypoxic survival and inhibition of pDC-differentiation and pyroptosis.Gene expression levels in CAL-1 cells are indicated by colors-red: elevated; green: downregulated; black: no alteration.

Author
Contributions: Conceptualization, S.N.; formal analysis, S.N., U.R., C.P. and C.M.; investigation, S.N.; data curation, C.P.; writing-original draft preparation, S.N.All authors have read and agreed to the published version of the manuscript.

Table 1 .
Target gene analysis for ONECUT2 based on RNA-sequencing data.