LXR agonist treatment of blastic plasmacytoid dendritic cell neoplasm restores cholesterol efflux and triggers apoptosis

Blastic plasmacytoid dendritic cell (PDC) neoplasm (BPDCN) is a rare aggressive malignancy derived from PDCs.¹  This disease is characterized by a heterogeneous presentation at diagnosis (from a disease limited to the skin to a leukemic syndrome with cytopenia and bone marrow involvement), clinical heterogeneity, and manifestations easily changing during disease progression.²  Currently, there is no consensus regarding the optimal treatment modality.²  Most BPDCN patients have a very aggressive clinical course with limited median overall survival.^2,3  It has been recently proposed that the frequent relapse after treatment and the poor prognosis can be related to the fact that the involvement of the central nervous system (CNS) is frequently undetected.⁴  Recently, BPDCN was classified by the World Health Organization (WHO) as a distinct entity in the group of “acute myeloid leukemia (AML) and related precursor neoplasms.”^2,5  Extensive characterization of this malignancy is still limited and diagnosis overlap may exist with immature AML, monoblastic and undifferentiated leukemia. Thus, a better understanding of this leukemia and new therapeutic approaches are urgently needed.

Previous studies have identified a cholesterol metabolism dysregulation in different malignant cells leading to intracellular cholesterol accumulation.^6,7  Cellular cholesterol content results from cholesterol uptake and biosynthesis through the mevalonate pathway, while its elimination is mediated by cholesterol efflux (Figure 1A). Cholesterol uptake involves plasma lipoproteins (mainly LDL and VLDL) after interactions with their specific receptors, LDLR and VLDLR, respectively. Cholesterol efflux implicates mainly adenosine triphosphate–binding cassettes (ABCs) A1 and G1 (ABCA1 and ABCG1, respectively) in association with extracellular cholesterol acceptors, including: apolipoprotein A1/E (APOA1 and APOE, respectively) or lipoprotein particles (eg, nascent high-density lipoprotein [HDL] or HDL2).⁸ 

Leukemic cells (AML and chronic myeloid leukemia) have been shown to increase LDLR expression,⁶  decrease LDLR degradation,⁷  and stimulate cholesterol biosynthesis resulting in cholesterol accumulation.⁶  Cholesterol regulates critical cellular functions, including plasma membrane formation, fluidity, and permeability.⁹  These latter functions are implicated in survival signaling pathway activation (eg, Akt)¹⁰  and proliferation.^11,12  For instance, stimulation of cholesterol efflux inhibits interleukin-3 (IL-3)-induced hematological progenitor cell proliferation.^13,14  Interestingly, BPDCN cells express high levels of IL-3 receptor α chain (CD123), and IL-3 is a BPDCN survival factor.^1,15  A targeted therapy directed against IL-3 receptor, called SL-401 associating IL-3 with the catalytic and translocation domains of diphteria toxin, has been tested in a phase 1/2 study with encouraging results.^16,17  Whether cholesterol homeostasis is dysregulated in BPDCN and contributes to its aggressiveness or determines response to therapies remains to be determined.

Cholesterol homeostasis is regulated at least by liver X receptors (LXRs). These nuclear receptors are expressed as 2 isoforms, with LXRβ being the ubiquitous isoform whereas LXRα expression is restricted to cells with high cholesterol turnover (eg, hepatocytes or macrophages).^18,19  The LXR pathway is activated by intermediates from the mevalonate pathway, endogenous oxidized cholesterol derivatives (called oxysterols), and synthetic agonists (eg, T0901317 or GW3965).^19,20  These synthetic compounds are of great interest for therapeutic use because LXR are considered as a promising target in different diseases.^19-21  LXR activation upregulates the expression of several genes involved in cholesterol homeostasis (called LXR target genes), including: ABCA1, ABCG1,²¹  and APOE (related to cholesterol efflux),²²  as well as the “inducible degrader of the low-density lipoprotein receptor” preventing cholesterol uptake through LDLR/VLDLR degradation.^23,24  Overall, these mechanisms participate in decreased intracellular cholesterol content. LXRs are functionally expressed in normal PDCs and in a leukemic PDC cell line,²⁵  but no data are available on the effects of LXR agonists on BPDCN.

The goal of this study was to determine whether BPDCN exhibit a specific gene signature based on genes involved in cholesterol efflux and uptake, in comparison with other leukemic cells (AML and T-acute lymphoblastic leukemia [T-ALL]) and normal PDCs. Because LXR activation controls cholesterol homeostasis via LXR target genes, we studied whether LXR agonist treatment stimulates cholesterol efflux. Effects of LXR activation on cell proliferation and survival were evaluated in vitro, using primary BPDCN samples and 2 established BPDCN cell lines (CAL-1 and GEN2.2). The in vivo LXR agonist therapeutic effect was evaluated using a BPDCN xenograft model treated with the T0901317 LXR agonist.

Twenty-three BPDCN samples were obtained at diagnosis (sample collection authorization number DC-2008-713). BPDCN was diagnosed based on histopathology and immunostaining of cutaneous lesions, blood, or bone marrow samples, as described.^26-28  This study was approved by our local ethics committee (Comité de Protection des Personnes [CPP] Est II, Besançon, France).

Two established BPDCN cell lines (CAL-1 and GEN2.2),^29,30  primary BPDCN cells isolated from a patient and expanded in NOD-SCID IL2Rγc-deficient (NSG) mice (The Jackson Laboratory, Sacramento, CA) (referred to hereafter as BES1), as well as 11 BPDCN samples with different BPDCN infiltration (supplemental Table 1, available on the Blood Web site) from newly diagnosed patients were used for in vitro assays. Culture of BPDCN cells and isolation of primary BPDCN samples are described in supplemental Methods.

The following samples were submitted to transcriptomic analysis using the GeneChip Human Genome U133 Plus 2.0 Array (Affymetrix, Santa Clara, CA):

• 12 BPDCN samples,

• 65 AML samples (including different French-American-British subtypes: 25 M0, 11 M1, 10 M2, 1 M3, 11 M4, 6 M5, and 1 M6) (Unité 837, Institut de Recherches sur le Cancer de Lille [IRCL], Lille, France),

• 35 T-ALL samples (Unité 1151, Assistance Publique–Hôpitaux de Paris [AP-HP], Hôpital Necker-Enfants Malades, Paris, France) (available at https://www.dropbox.com/sh/v21hg015hf515gw/AAC63OgjzcXXqTMmyca5sjVca?dl=0), and

• 5 primary PDCs (available on the Gene Expression Omnibus [GEO] database, under recording numbers: GSM705329, GSM705330, GSM705331, GSM705332, GSM705333).

Data were analyzed using dChip software (http://www.softpedia.com/get/Science-CAD/dChip.shtml) based on the expression of cholesterol homeostasis and LXR-related genes (LXRA, LXRB, ABCA1, ABCG1, APOE, SREBF1, FASN, LDLR, and VLDLR) plus the NFKB1 gene.

Transcription of LXR target genes (ABCA1, ABCG1) and genes coding proteins involved in the intrinsic apoptosis (BCL2, BAK1, BAX) was evaluated by quantitative reverse transcription polymerase chain reaction (qRT-PCR), as described.²⁵  Details are given in supplemental Methods.

CAL-1 cells were used to assess cholesterol efflux, as described in supplemental Methods and in Ishibashi et al.³¹ 

Cytotoxic effects of LXR agonists were evaluated by staining with Annexin V (AnxV) and 7-amino-actinomycin D (7AAD) (fluorescein isothiocyanate–conjugated AnxV/7AAD; BD Biosciences, Le Pont de Claix, France) or caspase-9 activation (CaspGLOW Fluorescein Active Caspase Staining kit; eBioscience), according to the manufacturer’s instructions. Proliferation was assessed on CAL-1 cells after labeling with the cell proliferation dye, eFluor 450 (eBioscience SA, Paris, France). BPDCN gating was performed using antibodies against CD123, CD131, and CD304 (supplemental Table 3). Cell cycle analysis was performed after cell fixation in ethanol 70% (overnight, 4°C), and by propidium iodide staining in a solution containing RNAse (overnight, 4°C). Cell fluorescence was evaluated using a CANTO II cytometer (BD Biosciences, San Jose, CA) and DIVA 6.2 software (BD Biosciences), except for cell cycle analysis where a FC500 cytometer with CXP and WinCycle softwares (Beckman Coulter Immunotech, Miami, FL) were used.

Whole-cell protein fraction was obtained by cell lysis in Laemmli buffer (supplemental Methods). Nuclear and cytosolic fractions were separated by cell lysis using a hypotonic, and then hypertonic, buffer solution (for osmosis restauration, supplemental Methods). Cytosolic fraction was isolated by centrifugation while nuclei were lysed in Laemmli buffer. Proteins were separated by electrophoresis on 8.5% or 12% sodium dodecyl sulfate–polyacrylamide gels and transferred to polyvinylidene difluoride membranes (GE Healthcare). Blots were then saturated with 5% milk before incubation with specific antibodies (supplemental Table 3). Blotted proteins were detected and quantified on a bioluminescence imager and BIO-1D advanced software (Vilber-Lourmat, Marne-la-Vallée, France) after blots were incubated with a horseradish peroxidase–conjugated appropriate secondary antibody. Details on blot saturation and quantification are given in supplemental Methods.

Protein expression (CD123 and ABCA1) and phosphorylation (p65, STAT5, and Akt) were evaluated by immunofluorescent staining, as previously described²⁵  (antibodies used appear in supplemental Table 3). Nuclei were labeled with 4′,6-diamidino-2-phenylindole (DAPI; Sigma-Aldrich) and cholesterol content was investigated using the free cholesterol marker filipin (Sigma-Aldrich), according to the manufacturer’s recommendations. Relative fluorescence intensity of filipin staining was measured with the ImageJ application, and determined as: corrected total cell fluorescence = [“integrated density” − (“area of selected cell” × “mean fluorescence of background readings”)]/untreated conditions.

NSG mice were irradiated (2 Gy), inoculated IV 18 hours later with 1 × 10⁶ CAL-1 cells, and treated intraperitoneally 7 days later with 6 injections of T0901317 (total experimental dose, 30 or 60 mg/kg, respectively) or with dimethyl sulfoxide (DMSO)/phosphate-buffered saline (PBS) control solution. Mouse monitoring and quantification of BPDCN cell infiltrate were described in supplemental Methods. Experimentation (#11007R) was approved by our local ethics committee (#58, approved by the French Ministry of Higher Education and Research) and conducted in accordance with the European Union Directive 2010/63.

Statistical analyses were performed by GraphPad Prism version 6 (GraphPad Software, San Diego, CA), using the Mann-Whitney, Wilcoxon, or Mantel-Cox test (*P < .05, **P < .01, ***P < .001, ****P < .0001). Data are expressed as mean ± standard error of the mean (SEM).

Twelve primary BPDCN samples were analyzed using the Affymetrix U133-2 messenger RNA (mRNA) microarray for the expression of cholesterol homeostasis and LXR-related genes (LXRA, LXRB, ABCA1, ABCG1, APOE, SREBF1, FASN, LDLR, and VLDLR) plus the NFKB1 gene. Comparison with AML and T-ALL samples revealed a specific BPDCN sample clustering, associated with a significant downregulation of LXR target genes ABCA1 and ABCG1 (associated with cholesterol efflux) and an upregulation of the VLDLR gene (linked to cholesterol entry). NFKB1 gene upregulation in BPDCN samples was confirmed (Figure 1B; supplemental Figure 1).³²  Comparison of the BPDCN samples with primary PDC samples showed a similar clustering associated with a significant downregulation of LXR target genes, SREBF1 and ABCG1, whereas VLDLR and NFKB1 genes were upregulated (Figure 1C; supplemental Figure 1). This shows that the BPDCN transcriptomic profile is independent of the PDC cell lineage. A significant downregulation of ABCA1 and ABCG1 gene transcription was confirmed by qRT-PCR analysis in 2 BPDCN cell lines (GEN2.2 and CAL-1), BES1 cells, and 4 primary BPDCN samples compared with nonleukemic PDC samples (Figure 1D). Overall, these data identify a specific perturbation of cholesterol homeostasis- and LXR-related gene transcription in BPDCN.

Treatment with 2 synthetic LXR agonists (T0901317 and GW3965, 1 µM, 24 hours) upregulated ABCA1 and ABCG1 gene transcription in CAL-1, GEN2.2 cell lines, and in 4 primary BPDCN samples (Figure 2A). ABCA1 and LXRα proteins were increased after LXR activation in CAL-1 cells, as assessed by western blot analysis (Figure 2B) and in CAL-1 and GEN2.2 cells, as assessed by confocal microscopy (supplemental Figure 2). Increase of ABCA1 expression after LXR activation was analyzed in 1 primary BPDCN sample by confocal microscopy (Figure 2C). Because LXR activation induces cholesterol efflux through ABCA1 and ABCG1 in cooperation with cholesterol acceptors, such as APOA1 and HDL2,^8,21  we then investigated cholesterol efflux using CAL-1 cells preloaded with ³H-cholesterol (1 µCi/mL, 24 hours), and treated with either T0901317 or GW3965 (1 µM, 24 hours) before the addition of APOA1 (10 µg/mL) or HDL2 (20 mg of protein per mL) cholesterol acceptors for 4 hours (Figure 2D). Radioactivity measurement in media and cells demonstrated that LXR agonist treatment significantly increased cholesterol efflux in both conditions (Figure 2E). Intracellular cholesterol staining revealed that LXR agonist treatment followed by APOA1 addition induced a significant diminution of total cholesterol content in CAL-1, GEN2.2 cells, and in 4 primary BPDCN samples (*P < .05), as assessed by confocal microscopy (Figure 2F). Overall, these data demonstrated that LXR agonist treatment of BPDCN stimulates cholesterol efflux via ABCA1 and ABCG1 transporters.

Because LXR activation regulates cell proliferation and survival,^13,20  we investigated the effects of LXR stimulation on BPDCN cells. CAL-1 cells were treated with increasing concentrations (1 µM, 5 µM, 10 µM) of LXR agonists for 24, 48, or 72 hours. Proliferation analysis of viable cells (AnxV⁻/7AAD⁻) demonstrated a significant decrease of cell proliferation induced by LXR agonist treatment (10 µM, 72 hours, P < .001), as assessed by cytometry (Figure 3A). Cell cycle phase analysis demonstrated a significant G1 phase retention associated with a diminution of cells in the S phase in a time-dependent (24 or 72 hours) manner (Figure 3B).

Exposure of CAL-1 and GEN2.2 cells to increasing concentrations of LXR agonists (10 µM, 30 µM, or 50 µM) demonstrated a significant cell death induction for concentrations higher than 10 µM, as assessed by AnxV/7AAD staining and cytometry. The cytotoxic effect of LXR agonists was confirmed in BES1 cells and 5 primary BPDCN samples of 5 tested (P < .05; Figure 4A-B). Assessment of viable BPDCN cells (AnxV⁻/7AAD⁻/CD123⁺/CD304⁺) in 1 blood sample from a BPDCN patient revealed after treatment with LXR agonists, a preferential decrease of viable BPDCN cells (77% vs 50% and 30%, for vehicle- vs T0901317- and GW3965-treated samples, respectively) (Figure 4B). This suggests a specific cytotoxic effect of LXR agonists on BPDCN cells. Western blot analysis of CAL-1 cells treated with LXR agonists for 6 hours showed a caspase-3 and caspase-9 cleavage, suggesting apoptosis induction (Figure 4C). This was confirmed in CAL-1 and GEN2.2 cells by nucleus fragmentation induced by LXR agonists and at morphological levels, as assessed by confocal microscopy (supplemental Figure 3A-B). Caspase-9 activation was confirmed after LXR agonist treatment in GEN2.2 cells and 2 primary BPDCN cells, as assessed by cytometry (Figure 4D). BAX- and BAK1-coding gene upregulation in CAL-1 and GEN2.2 cells after LXR activation was detected (Figure 4E). This suggests the involvement of the intrinsic apoptosis. Overall, this indicated that LXR agonist treatment stimulates apoptotic cell death in BPDCN.

Because LXR activation inhibits BPDCN survival, we wondered whether LXR stimulation would interfere with the following survival signaling pathways, IL-3 and NF-κB. IL-3 was described to induce STAT5 and Akt activation, both involved in leukemic cell survival.^33-36  NF-κB activation was reported to maintain BPDCN cell survival.³²  To investigate the effects of LXR activation on IL-3–induced STAT5 and Akt activation, CAL-1 cells were treated with increasing noncytotoxic concentrations of LXR agonists (1-10 µM, 24 hours), followed by IL-3 stimulation (10 ng/mL, 30 minutes). Western blot analysis demonstrated a sustained diminution of STAT5 and Akt phosphorylation induced by LXR agonist treatment (Figure 5A). These effects were confirmed by confocal microscopy in 3 primary BPDCN samples (Figure 5B), as well as in CAL-1 and GEN2.2 cells (supplemental Figure 4A).

To demonstrate the involvement of constitutive NF-κB activation in BPDCN cell survival, CAL-1 and GEN2.2 cells were treated with increasing concentrations (12.5-100 µM, 24 hours) of the NF-κB p65 inhibitor, JSH23. A significant increase of BPDCN cell death was revealed by AnxV/7AAD staining (**P < .01, Figure 5C). LXR agonists decreased NF-κB p65 phosphorylation in 4 primary BPDCN samples (Figure 5D) and in CAL-1, GEN2.2 cells (supplemental Figure 4B), as assessed by confocal microscopy. Western blot analysis of CAL-1 cells pretreated with LXR agonists (1 µM, 24 hours) demonstrated an inhibition of p50, p65, and c-Rel NF-κB subunit nuclear translocation, induced by a NF-κB activator (R848, 1 µg/mL, 6 hours) (Figure 5E). Overall, these data demonstrated that LXR stimulation in BPDCN cells inhibits IL-3–induced STAT5 and Akt activation, as well as NF-κB activation at phosphorylation and nuclear translocation levels. This may contribute to the cytotoxic effects of LXR agonist treatment on BPDCN.

Cholesterol efflux through ABCA1/ABCG1 inhibits IL-3–induced hematopoietic stem cell (HSC) proliferation,^13,14  and LXR activation in BPDCN interferes with the IL-3 signaling pathway. To investigate the contribution of LXR-stimulated cholesterol efflux in these effects, CAL-1 and GEN2.2 cells or a primary BPDCN sample were treated with LXR agonists (1 µM, 24 hours), then with APOA1 (10 µg/mL, 4 hours) followed by IL-3 (10 ng/mL, 30 minutes). Addition of APOA1 markedly diminished IL-3–induced STAT5 and Akt phosphorylation in all LXR-treated BPDCN cells (Figure 6A-B). Treatment of CAL-1 cells with increasing noncytotoxic concentrations of LXR agonists (5-10 µM) in the presence of APOA1 (0-20 µg/mL) showed a significant increase of dead BPDCN cells (Figure 6C). Overall, these data demonstrated that cholesterol efflux increases LXR agonist-mediated effects. To go further on cholesterol dependency of BPDCN, cholesterol was deprived from BPDCN cells by using either an inhibitor of the mevalonate pathway, atorvastatin, or a compound inducing cholesterol removal from cells, methyl-β-cyclodextrin.³⁷  Cell death analysis 24 hours later by AnxV/7AAD staining and cytometry demonstrated a significant BPDCN cell death (supplemental Figure 5).

To assess LXR therapeutic effects in vivo, sublethally irradiated (2 Gy) NSG mice were grafted with 1 million CAL-1 cells. After 7 days, mice were treated with 2 doses of LXR agonist (T0901317 30 mg/kg or 60 mg/kg) or vehicle, every 2 days until sacrifice. Although CAL-1 cell injection induced a significant persistent decrease of red blood cell (RBC), platelet counts, as well as hemoglobin (Hb) concentration, T0901317-treated mice showed a significant prevention of cytopenia, including RBC counts and Hb concentration (Figure 7A). At sacrifice, a diminution of CAL-1 cell-induced splenomegaly was observed (Figure 7B), supported by a potent decrease of spleen and bone marrow BPDCN cell infiltration, as assessed by flow cytometry (Figure 7C). In an additional experiment, treatment with T0901317 (30 mg/kg or 60 mg/kg) significantly increased the overall survival of NSG mice inoculated with CAL-1 cells compared with CAL-1-inoculated and vehicle-treated mice (Figure 7D-E). Overall, these data provide the in vitro and in vivo demonstration of a therapeutic effect of LXR agonist on BPDCN through different mechanisms, including signaling pathway regulation and cholesterol efflux.

Cholesterol is a critical component for cell growth and proliferation, as illustrated by the inhibition of HSC proliferation occurring after cholesterol efflux through ABC transporters.^13,14  In physiological conditions, cholesterol homeostasis is tightly controlled by the LXR signaling pathway. Here, we identified that the aggressive hematological malignancy BPDCN exhibits a specific transcriptomic signature with a downregulation of several LXR target genes involved in cholesterol homeostasis. This may lead to cholesterol accumulation within leukemic cells, responsible for high proliferative properties. We reported that LXR agonist treatment increases LXR target gene expression in BPDCN, stimulates cholesterol efflux from these cells, and is associated with the inhibition of proliferation and survival. These 2 effects may result from the interference of LXR activation with 2 BPDCN survival/proliferative pathways, namely IL-3 and NF-κB signaling pathways. The effects of LXR agonists are amplified by the addition of the lipid acceptor APOA1 known to enhance cholesterol efflux. All of these effects (except cell proliferation) were assessed in vitro in 2 BPDCN cell lines, expanded primary BPDCN cells, and several primary BPDCN samples isolated from 11 different patients. An in vivo therapeutic effect of LXR agonist is also observed in a xenograft model with reduction of BPDCN cell infiltration, prevention of BPDCN-induced cytopenia, and increased mouse survival. These data highlight an unrevealed perturbation of cholesterol homeostasis and LXR activity in BPDCN, and identify a new approach based on LXR agonists to treat this aggressive hematological malignancy.

Treatment of BPDCN with LXR agonists has several effects depending on their concentration. Restoration of LXR target gene expression (inducing cholesterol efflux) and inhibition of IL-3 and NF-κB signaling pathways occurred for concentrations between 1 and 10 µM (ranges usually used in malignant cells^10,38-42 ). BPDCN cell proliferation (assessed with the proliferative BPDCN cell line, CAL-1) was significantly inhibited for 10 µM, as reported in several hematological malignancies and solid tumors.^39,42,43  Concentrations higher than 10 µM induced a significant BPDCN cell death, as reported for ovarian, breast, and colon cancer cells.^43-45  All of these data validate the concentrations of LXR agonists used in this study.

LXR activation in BPDCN triggers an apoptotic cell death mechanism based on different features, namely: exposure of phosphatidylserine (assessed by AnxV staining), cleavage of caspase-3 and caspase-9, as well as nucleus fragmentation. Induction of intrinsic apoptosis by LXR agonists was previously described in ovarian carcinoma cells,⁴⁵  and confirmed here associated with a significant increase of BAX and BAK1 transcripts (encoding proapoptotic proteins). LXR-induced cell death in breast cancer cells has been demonstrated to implicate BAX upregulation and to be dependent on cholesterol efflux through ABCG1.⁴³  In our study, a stimulation of ABCA1/ABCG1-dependent cholesterol efflux via the addition of APOA1 increases LXR-mediated cytotoxic effects. APOA1-stimulated cholesterol efflux also potentiates the LXR-induced inhibition of Akt and STAT5 phosphorylation. This is in line with a previous report in prostate cancer cells showing Akt inhibition by increased cholesterol efflux.¹⁰  Although exogenous APOA1 supplementation is required for in vitro assays with BPDCN cells, circulating lipid-free APOA1 (mainly produced by liver and intestine),⁸  or APOA1 present in nascent HDL is likely to potentiate the impact of LXR agonist treatment in vivo. Overall, LXR stimulation in BPDCN exerts an antileukemic effect that can be enhanced by increasing cholesterol efflux.

Therapeutic strategies of BPDCN propose to interfere with IL-3^2,16,34,46  because BPDCNs express high levels of CD123,^2,26  and IL-3 is a BPDCN survival factor.^1,15  Here, we explored how LXR activation and LXR-induced cholesterol efflux interact with this pathway. No modification of CD131 and CD123 expression on BPDCN cells after LXR activation is observed (data not shown), whereas LXR agonists inhibit IL-3–induced Akt and STAT5 phosphorylation in BPDCN. Western blot analysis demonstrated that LXR agonists had no effect on STAT5 and Akt protein expression, suggesting a predominant effect at the phosphorylation levels.

BPDCN samples exhibit a specific downregulation of LXR target genes ABCG1, ABCA1, and SREBF1, in favor of repression of LXR transcriptomic activity. However, although LXR target gene transcription is decreased in BPDCN, the transcripts of the 2 LXR coding genes (LXRA and LXRB) are not and basal LXRβ protein (supplemental Figure 2B) is present. This suggests posttranscriptional regulation of LXR activity in BPDCN. LXR activity in normal PDC is inhibited by prior NF-κB activation.²⁵  Because NF-κB is constitutively activated in BPDCN,³²  and NF-κB p105 precursor-coding gene (NFKB1) is upregulated, LXR repression in BPDCN may be related to NF-κB activation. Restoration of the LXR pathway by agonist treatment in BPDCN inhibits the constitutive NF-κB activation at 3 different levels: p65 phosphorylation, nuclear translocation of the p50, p65, and cRel subunits, as well as transcription of NFKB1 (data not shown). Our study confirms constitutive NF-κB activation in BPDCN cells,³²  and demonstrates that inhibition of p65 translocation by JSH-23 is sufficient to induce BPDCN cell death. This suggests that LXR agonist-induced cell death is related to NF-κB inhibition. Interestingly, STAT5 and NF-κB were also reported to promote G1 to S cell cycle phase transition through cyclin D1 induction, and thus cell proliferation.^47,48  In our study, BPDCN cells are retained in G1 phase after LXR activation. This suggests that LXR-induced STAT5 and NF-κB inhibition can be involved in both the inhibition of cell proliferation (through cell cycle arrest) and BPDCN cell death.

LXR agonist treatment of mice grafted with BPDCN cells prevents leukemia-induced cytopenia, reduces BPDCN spleen and bone marrow infiltrations, and slightly but significantly improves mouse survival. To date, the development of LXR agonists in clinical settings has been hampered by unwanted systemic side effects, such as fatty liver disease and LDL elevation.¹⁹  Synthetic LXR agonists have been shown to exert therapeutic effects in mouse models of Alzheimer disease after oral administration.¹⁹  This suggests that these agonists can cross the blood-brain barrier and may target BPDCN cells infiltrating the CNS. The CNS may represent a blast cell sanctuary in BPDCN patients with leukemic presentation both at diagnosis and at relapse.⁴  Efforts are currently being made to generate new synthetic agonists with increased specificity for the LXRβ isoform, expressed by BPDCNs, to limit steatosis^19,49  and/or to stimulate LXR specifically in a target tissue.^50,51  Our study supports a new approach for BPDCN treatment using these new synthetic LXR agonists.

The authors thank T. Maeda (Nagasaki University, Japan) for kindly providing the BPDCN cell line CAL-1, J. Plumas and L. Chaperot for kindly providing the BPDCN cell line GEN2.2, W. Le Goff (INSERM UMR S1166, University “Pierre et Marie Curie”, Paris, France) for helpful discussions on cholesterol metabolism, the Cytology laboratory of the Etablissement Français du Sang Bourgogne Franche-Comté (BFC), for blood cell and platelet counts, Sarah Odrion, and Alexis Varin for editorial assistance.

This work was supported by the Etablissement Français du Sang (grant 2011-11) (P.S.), the Agence Nationale de la Recherche (LabEx LipSTIC, ANR-11-LABX-0021), the Conseil Régional de Franche-Comté (soutien au LabEx LipSTIC 2015 and 2016) (P.S.), and the Fondation de Coopération Scientifique Bourgogne Franche-Comté (A.C., D.M., and P.S. via the Bonus Qualité Recherche BFC).

There are no current or pending patents related to this work. Two material transfer agreements related to the use of CAL-1 cell line and the GEN2.2 cell line are available.

Contribution: A.C. performed most of the experiments, collected, assembled, and analyzed data, performed statistical analysis, and wrote the manuscript; D.M. performed cholesterol efflux experiments and helped to write the manuscript; A.R. and C.R. performed transcriptomic experiments and analysis; C.C. performed cell death analysis by flow cytometry, some immunoblotting, and qRT-PCR experiments; T.G. performed some confocal microscopy experiments; L.P. performed cell cycle experiments; B.L., F.A.-D., and F.B. performed and supported in vivo experiments; S.P. commented on the manuscript and helped to write it; S.B., C.P., and E.M. provided leukemia samples and diagnosis; L.L. commented on the manuscript and provided major funding support; F.G.-O. provided study material, collected data, and helped to write the manuscript; and P.S. supervised research, analyzed data, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

The current affiliation for B.L. is Université Pierre et Marie Curie, UMR 938, INSERM, Paris, France.

Correspondence: Philippe Saas, EFS BFC, UMR 1098, INSERM, 8 rue JFX Girod, BP1937, F-25020 Besançon Cedex, France; e-mail: philippe.saas@efs.sante.fr.