Blastic Plasmacytoid Dendritic Cell Neoplasm Is Dependent on BCL2 and Sensitive to Venetoclax

Blastic plasmacytoid dendritic cell neoplasm (BPDCN) is an aggressive hematologic malignancy that presents with skin nodules and tumors, lymph node and splenic enlargement, central nervous system involvement, circulating leukemia, and/or bone marrow infiltration (1). There is no standard therapy, and patients empirically receive chemotherapy regimens based on retrospective case series (2). For patients who respond to chemotherapy, autologous or allogeneic stem cell transplantation may prolong survival (3). However, median survival is only approximately 1 year, or even shorter, in patients who have disseminated disease (1, 2).

BPDCN has had many prior names, including CD4⁺CD56⁺ hematodermic neoplasm, blastic natural killer cell lymphoma, and agranular CD4⁺ natural killer cell leukemia, which has complicated study of the disease (1). In 2008, BPDCN was recognized as the malignant counterpart of plasmacytoid dendritic cells (pDC; ref. 4). However, the pathogenesis underlying the transformation of hematopoietic progenitors to BPDCN remains unclear. Targeted sequencing identified mutated genes in BPDCN that overlap with those observed in other hematologic malignancies, particularly myeloid diseases, including TET2, TP53, ASXL1, and RNA splicing factors (5, 6). However, no BPDCN-specific mutations or immediately targetable alterations have been reported from those studies.

A gene expression analysis suggested that BPDCN was similar to acute myeloid leukemia (AML), and identified the antiapoptotic gene BCL2 as more highly expressed in BPDCN compared with normal pDCs (7). BCL2 is also expressed in AML, and some AMLs are sensitive to the BCL2 inhibitor venetoclax (8). Venetoclax has a favorable safety profile and was recently approved by the FDA for use in patients with relapsed chronic lymphocytic leukemia (CLL; ref. 9). Here, we demonstrate that BPDCN is dependent on BCL2 and that venetoclax is active in BPDCN, including in 2 patients with relapsed/refractory disease. Thus, BCL2 inhibition could provide a novel therapeutic strategy in this chemotherapy-resistant disease.

We performed immunohistochemistry on BPDCN biopsies from bone marrow or skin, and in all cases, we observed prominent BCL2 staining compared with surrounding normal tissue (Fig. 1A; Supplementary Fig. S1). We saw similar BCL2 expression in AML blasts from bone marrow and leukemia cutis (Supplementary Fig. S1). These data suggested that, like AML, BPDCN might have some degree of BCL2 dependence. However, cell death in response to BCL2 inhibition depends on a complex balance of pro- and antiapoptotic factors, which makes sensitivity prediction challenging based on expression alone (10).

Therefore, we performed functional mitochondrial profiling and conventional cytotoxicity assays to test BCL2 inhibition in BPDCN. We first compared a BPDCN cell line, CAL1 (11), with a series of AML cell lines that have a range of sensitivity to BCL2 inhibition (12). CAL1 cells expressed similar levels of BCL2 protein to the highest BCL2-expressing AML cells (Fig. 1B), but there was variability in the abundance of other apoptotic pathway proteins. To assess the functional dependence of these cell lines on specific BCL2 family members, we performed BH3 profiling. This technique exposes mitochondria to peptides that mimic the BH3 domain of proapoptotic BCL2 family members and measures the induced change in mitochondrial permeability, the “point of no return” for apoptotic cell death (13).

We determined the propensity of each cell line to initiate mitochondrial apoptosis after stimulation with a nonspecific prodeath BIM peptide (overall priming), and the relative dependency on three antiapoptotic BCL2 family members: BCL2, BCLXL, and MCL1. CAL1 cells were most dependent on BCL2, as they showed significant priming in response to BAD stimulation (BCL2/BCLXL), but only minor priming after HRK (BCLXL only) and none after MS1 (MCL1) stimulation (Fig. 1C). In contrast, AML cells were equally or less BCL2 dependent, and/or were codependent on BCLXL or MCL1 in addition to BCL2.

To directly measure responses to pharmacologic BCL2 inhibition, we treated cells with venetoclax, a BH3 mimetic molecule that displaces proapoptotic proteins such as BIM from sequestration by BCL2, allowing them to initiate mitochondrial permeabilization (14). We performed dynamic BH3 profiling (15), which measures the increase in apoptotic priming induced by incubation with a drug. Dynamic BH3 profiling strongly correlates with eventual induction of apoptotic cell death by the same treatment yet requires only short-term exposure to drug (≤4 hours), which is advantageous in analyses of primary cells. By measuring the dose-dependent cytochrome c release induced by the BIM peptide, we calculated the change in overall apoptotic priming caused by venetoclax pretreatment (“delta priming”). CAL1 cells had an equal or higher delta priming compared with the AML cells tested (Fig. 1D).

As predicted by this result, CAL1 cells were equally or more sensitive to venetoclax compared with AML cells in viability assays (Fig. 1E). We confirmed that venetoclax induces dose-dependent apoptotic cell death in CAL1 cells by measuring Annexin V and propidium iodide staining following treatment (Supplementary Fig. S2A and S2B). We also noted that dynamic BH3 profiling significantly correlated with Annexin V positivity in response to venetoclax (P = 0.0067; Fig. 1F), supporting the validity of this assay as a functional surrogate of drug sensitivity. The predictions made by BH3 profiling across all cell lines were also confirmed using navitoclax (ABT-263), which inhibits both BCL2 and BCLXL, and A-1331852, which targets BCLXL only (ref. 16; Supplementary Fig. S3). Together, these data suggest that BPDCN cells are sensitive to venetoclax, at least in part because they are highly dependent on BCL2 to inhibit mitochondrial apoptosis.

To test this hypothesis in primary cells, we analyzed patient bone marrow aspirates and skin biopsies involved by BPDCN, as well as BPDCN patient–derived xenografts (PDX; ref. 17). PDXs displayed pathologic characteristics of human BPDCN, including infiltration of the bone marrow and spleen by CD4⁺CD56⁺CD123⁺ leukemia cells that were also BCL2-positive (Supplementary Fig. S4). Patient tumor samples and PDXs were subjected to targeted DNA sequencing and represent a variety of genotypes (Supplementary Table S1).

Culturing primary leukemia cells in vitro to perform drug treatment assays is challenging. Ex vivo cytotoxicity assays with primary BPDCNs suggested a dose-dependent response to venetoclax in some cases (Supplementary Fig. S5A and S5B), but maintaining viability even in vehicle-treated samples for more than 8 hours was not uniformly feasible. Therefore, we performed BH3 profiling in primary BPDCNs similarly to how we had done in cell lines. In addition, we used flow cytometry–based BH3 profiling to measure cytochrome c release in defined subpopulations of interest (18). Using this technique, we were able to selectively assess responses in leukemia cells admixed with normal cells. For comparison, we also tested a randomly selected set of primary AMLs. Baseline BCL2/BCLXL dependency was higher in BPDCN compared with AML (69.1% vs. 2.1% priming by BAD peptide, P < 0.0001 by t test; Fig. 2A). Similarly, when permeabilized cells were directly stimulated with equivalent doses of venetoclax, BPDCNs had a higher level of cytochrome c release than the AMLs tested (57.4% vs. 2.62%, P = 0.0001; Fig. 2B).

Next, we used dynamic BH3 profiling to measure the change in apoptotic priming in response to venetoclax treatment of live cells with intact membranes. All BPDCNs had a significant increase in priming, or a decrease in their apoptotic threshold, after short-term treatment with venetoclax, and BPDCN priming was higher than in AML (59.8% vs. 12.1%, P < 0.0001; Fig. 2C). Normal bone marrow had significantly less priming than BPDCN in response to the same dose of venetoclax (18.5% vs. 59.8%, P = 0.0002; Fig. 2C), suggesting a therapeutic window.

We performed an orthogonal assay for apoptotic cell death by measuring cell surface Annexin V in primary BPDCN and AML after short-term venetoclax treatment. On average, BPDCN had a greater increase in Annexin V positivity than AML (2.25-fold vs. 1.2-fold, P = 0.03; Fig. 2D). Annexin V positivity was significantly correlated with the increase in priming measured by dynamic BH3 profiling (P = 0.0088; Fig. 2D). Together, these data suggest that BPDCN is highly BCL2 dependent and predicts that BPDCN may be sensitive to venetoclax therapy.

To test BPDCN response to BCL2 inhibition in vivo, we transplanted two PDXs into recipient animals. When human BPDCN was detectable in blood, we randomized animals to treatment with venetoclax or vehicle for 28 days. After 21 days, we sacrificed a subset for pharmacodynamic evaluation. Venetoclax treatment resulted in a reduced burden of human BPDCN cells in peripheral blood, bone marrow, and spleen (Fig. 3A and B; Supplementary Fig. S6). Histologic and immunohistochemical analyses confirmed decreased BPDCN in tissues and showed restoration of normal hematopoietic elements in treated animals (Fig. 3C). Animals receiving venetoclax had prolonged overall survival compared with vehicle-treated mice (median survival, 57 vs. 36 days, P = 0.0025; Fig. 3D).

We identified 2 patients with relapsed/refractory BPDCN who received therapy with venetoclax, prescribed off-label after the recent FDA approval in CLL. Both patients had no alternative therapeutic options, and their treating physicians prescribed venetoclax because of the reports of BPDCN responding to “lymphoid-like” leukemia regimens (2) and because of venetoclax's single-agent activity in AML (8). Patient #1 was an 80-year-old male who had been diagnosed with BPDCN 18 months prior and had disease involving bone marrow, lymph nodes, and diffuse cutaneous plaques and tumors. His BPDCN harbored the following mutations: ASXL1^Y591fs*, GNB1^K57E, IDH2^R140W, and NRAS^G12D. He had previously responded to and then progressed on conventional chemotherapy (doxorubicin, vincristine, and prednisone), an IDH2-targeting agent, and an IL3 receptor–targeting agent. Patient #2 was a 73-year-old male diagnosed with BPDCN 15 months prior, with skin, marrow, and widespread nodal involvement. His BPDCN harbored MPL^Y591N, TET2^M1456fs*, and TET2^Q1654fs* mutations. He had previously received and then relapsed after chemotherapy (hyperfractionated cyclophosphamide, vincristine, doxorubicin, and dexamethasone), an autologous stem cell transplant, an IL3 receptor–targeting agent, and decitabine.

We had previously performed dynamic BH3 profiling on biopsies of skin (Patient #1) and bone marrow (Patient #2), which predicted that their disease would respond to venetoclax (BPDCN3 and 6; Fig. 2B). Patient #1 received venetoclax orally with a weekly dose escalation of 20 mg, 50 mg, 100 mg, 200 mg, and finally 400 mg per day. He did not have any evidence of tumor lysis syndrome, despite rapid regression of visible tumors. At 4 weeks, his disease was restaged, which revealed a marked response in multiple cutaneous sites of involvement (Fig. 4A). His modified Severity Weighted Assessment Tool (mSWAT) skin scores decreased from 16.5 to 6.5. He had a decrease in the size of palpable cervical and preauricular lymph nodes, but did not have an appreciable response in his bone marrow at that time. Unfortunately, he died 2 weeks later with an intracranial hemorrhage before additional response evaluation could be performed. His platelets had been persistently below 10,000/μL despite transfusion, predating venetoclax treatment and likely related to bone marrow infiltration by BPDCN.

Patient #2 received a daily dose escalation of venetoclax of 50 mg, 100 mg, 200 mg, to a final dose of 400 mg daily. He did not have any evidence of tumor lysis syndrome nor any other toxicity related to venetoclax. At 4 weeks, his skin disease had substantially improved, and a PET-CT scan demonstrated a significant decrease in multistation lymphadenopathy and near-complete resolution of all areas of FDG-avidity (Fig. 4B and C). His bone marrow BPDCN blast count decreased from 85% before treatment to 44% after 6 weeks. He remained on venetoclax 400 mg daily for approximately 12 weeks, at which time he experienced disease progression.

We have demonstrated that BPDCN is dependent on BCL2 and is markedly sensitive to BCL2 inhibition with venetoclax. Amongst the primary leukemias, PDXs, and cell lines we tested, BPDCN was at least as sensitive to BCL2 inhibition compared with AML. The BPDCN response to venetoclax compares favorably with CLL, a disease that has shown remarkable single-agent activity, including in relapsed patients with unfavorable genetics (9). BCL2 inhibition should be formally evaluated in BPDCN as soon as possible, because there are few therapeutic options for these patients.

The mechanisms underlying BCL2 dependence in BPDCN are unclear. There are no recurrent DNA copy-number changes nor rearrangements reported involving the BCL2 locus (19, 20), although additional analysis of BPDCN genomics may provide more insight into its BCL2 dependency. Prior studies in AML have suggested that specific somatic mutations may be enriched in venetoclax-sensitive leukemias (8, 21). Larger numbers of patients with BPDCN treated with venetoclax with complete genetic annotation will be required to make definitive genotype–phenotype correlations in this disease. However, we note that the primary BPDCNs tested here harbored a variety of mutations in genes associated with hematologic malignancies (Supplementary Table S1), and all responded to venetoclax by BH3 profiling. Normal pDCs are selectively depleted in vivo by venetoclax compared with conventional dendritic cells and other hematopoietic cell types (22), suggesting that the sensitivity we observed could represent a lineage-specific dependency.

Finally, these data demonstrate precision cancer therapy directed by functional rather than genetic assessment. BH3 profiling served as a biomarker to identify BPDCN as dependent on BCL2 and likely to respond to venetoclax. The same technique could be used in trials (in BPDCN or other cancers) evaluating venetoclax to determine if drug-resistant cells switch their dependency to another antiapoptotic protein, which itself might be targetable. Furthermore, microenvironmental signals modulate apoptotic dependencies (22, 23), which suggests that correlative studies in clinical trials could analyze BPDCN from distinct anatomical sites to elucidate additional mechanisms of response and resistance. Finally, combination of venetoclax with other agents should also be evaluated in BPDCN, given that BCL2 inhibition may be synergistic with chemotherapy (24).

CAL1 cells were obtained from Takahiro Maeda (Nagasaki University) in 2012 (11). AML cell lines were obtained from the ATCC or DSMZ between 2008 and 2014. They were validated by short-tandem repeat profiling in 2014 prior to their use in these experiments, and they underwent Mycoplasma testing every 6 months.

Primary BPDCN and AML cells were collected from patients who had consented to Institutional Review Board (IRB)–approved research protocols for sample analysis from patients with hematologic malignancies. Bone marrow aspirate mononuclear cells from patients with BPDCN or AML were purified by Ficoll density centrifugation using standard procedures. Skin biopsies were exposed to an enzymatic digestion solution in 2.5 mL of DMEM/F12 media containing 125 U DNAse I (Sigma Aldrich #DN25), 100 U hyaluronidase (Sigma Aldrich #H3506), and 300 U collagenase IV (Gibco #17104-019). The tissue suspension was processed using gentleMACS Dissociator (Miltenyl Biotec) using the hTUMOR 1 program. The suspension was then incubated at 37°C for 30 minutes with constant agitation. Then, dissociation was repeated using the hTUMOR 1 program, and the 30-minute incubation was repeated. We then filtered the suspension through a 70-μm filter, and cells were centrifuged at 400 × g for 5 minutes. To lyse residual red blood cells, 100 μL of ice-cold water was added to the pellet for 15 seconds and then diluted to 50 mL with PBS. Then, cells were centrifuged and resuspended in RPMI media for subsequent analysis.

Samples for Western blotting were prepared by lysing 10⁶ cells in radioimmunoprecipitation assay buffer (Boston Bioproducts, #BP-115) containing 1x protease inhibitor (ThermoFisher, cat. 87786). The antibodies used in were BCL2 (BD Pharmingen; #551107), BCLXL (Cell Signaling Technologies; #2764), MCL1 (Cell Signaling Technologies; #5453), BIM (Cell Signaling Technologies; #2933), BAX (#2774), and β-actin (Sigma-Aldrich; #A5441). All antibodies were used at 1:1,000 dilution, except β-actin at 1:10,000. Western blots were imaged using an ImageQuant LAS-4000 (GE Healthcare; #28-9558-10).

Cell lines cultured in RPMI-1640 with 10% FBS were plated in a 96-well dish at a concentration of 5,000 cells per 120 μL media. Compounds were added to the cells in serial 5-fold dilutions. After a 72-hour incubation at 37°C, viability was determined by an MTT assay. In brief, 40 μL of 5 mg/mL MTT (EMD Millipore; #475989) was added to each well and incubated for 2 hours, and then 100 μL MTT lysis buffer was added to each well followed by additional 4-hour incubation. Absorbance values were measured using a SpectraMax M3 plate reader (Molecular Devices) at 570 and 630 nm. Viability curve values were generated using the nonlinear regression (curve fit) function in GraphPad Prism (GraphPad Software, Inc.).

The cells were treated with drug for the indicated times, washed with Annexin V binding buffer (ABB), and stained with 1:100 dilutions of Annexin V–FITC (Biolegend; #640906), CD45-APC (BD Biosciences; #340943), and CD123-PerCP-Cy5.5 (BD Pharmingen; #560904) for 30 minutes. Samples were then washed twice with ABB, propidium iodide (BD Pharmingen; #51-66211E) was added after the last wash, and then cells were analyzed using a Cytoflex flow cytometer (Beckman Coulter; #B53012).

Baseline and dynamic BH3 profiling were performed as described (13, 15). Dynamic BH3 profiling was performed after 4 hours of incubation of cell lines or primary leukemia cells in vehicle or 100 nmol/L venetoclax, unless otherwise specified. We used a flow cytometry–based BH3 profiling to perform the analysis, as previously described (18), using Zombie Aqua Dye (Biolegend; #423101) for viability, CD123-PerCP-Cy5.5 (BD Pharmingen; #560904), CD56-PECy7 (BD Pharmingen; #557747), and cytochrome c-Alexa Fluor 647 (Biolegend; #612310).

All animal experiments were approved by Institutional Animal Care and Use Committees. BPDCN PDXs were generated as described in the Dana-Farber Cancer Institute Public Repository of Xenografts (PRoXe.org; ref. 17). For the treatment trial, one million cells from two independent PDXs (PDX1 and PDX4) were injected into 16 NSG-SGM3 mice each [NOD/SCID IL2Rgnull-3/GM/SF, NOD.Cg-Prkdc^scidIl2rg^tm1WjlTg(CMV-IL3,CSF2,KITLG)1Eav/MloySzJ; Jackson Labs, #013062]. Mice were monitored weekly for evidence of human BPDCN in the peripheral blood. When peripheral human CD45⁺CD123⁺ cells were more than 1% of total white blood cells, animals were randomized to two groups, and treatment was started with vehicle or venetoclax 100 mg/kg/daily by oral gavage for 28 days. Three mice in each group were sacrificed at day 21 of treatment for pharmacodynamic assessment. The remaining animals were followed for survival. Both PDXs showed evidence of tumor response. PD and survival analysis for PDX4 are shown in Fig. 3, and PDX1 in Supplementary Fig. S4. Kaplan–Meier curves were compared using the log-rank test in GraphPad Prism. Immunohistochemistry was performed in the Dana-Farber/Harvard Cancer Center Specialized Histopathology Core Laboratory using standard protocols.

Patient samples were sequenced using a 95-gene targeted sequencing panel covering genes recurrently mutated in hematologic malignancies (25).

Patients with relapsed/refractory BPDCN who had exhausted other treatment options were prescribed venetoclax off-label by their treating physicians after they provided written informed consent. Prescriptions were filled in standard outpatient pharmacies after the FDA approval of venetoclax in 2016. The patients provided written informed consent to IRB-approved protocols for sample collection from patients with hematologic malignancies. They also signed additional specific consents for photographic data to be published, and all studies were in accordance with the Declaration of Helsinki. Patient #1 received oral venetoclax with a weekly dose escalation, starting at 20 mg daily for 7 days, followed by 50 mg daily for 7 days, 100 mg daily for 7 days, 200 mg daily for 7 days, followed by 400 mg daily. He was admitted to the hospital for the first dose initiation and the escalation to 50 mg to monitor for tumor lysis syndrome. Patient #2 received oral venetoclax with a daily dose escalation of 50 mg for one day, 100 mg for one day, 200 mg for one day, and then 400 mg daily. He was admitted to the hospital during the dose escalation to monitor for tumor lysis.

M.S. Davids is a consultant/advisory board member for AbbVie and Genentech. R.M. Stone is a consultant/advisory board member for AbbVie. M. Konopleva reports receiving a commercial research grant from AbbVie, and is a consultant/advisory board member for the same. N. Pemmaraju is a consultant/advisory board member for Stemline and LFB. A. Letai reports receiving commercial research grant from AbbVie, and is a consultant/advisory board member for the same. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Montero, L. Cabal-Hierro, M. Konopleva, N. Pemmaraju, A. Letai, A.A. Lane

Development of methodology: J. Montero, J. Stephansky, L. Cabal-Hierro, A. Letai, A.A. Lane

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Montero, J. Stephansky, T. Cai, G.K. Griffin, L.J. Hogdal, I. Galinsky, J.C. Aster, M.S. Davids, N.R. LeBoeuf, R.M. Stone, N. Pemmaraju, A.A. Lane

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Montero, J. Stephansky, T. Cai, G.K. Griffin, L. Cabal-Hierro, I. Galinsky, M.S. Davids, M. Konopleva, A. Letai, A.A. Lane

Writing, review, and/or revision of the manuscript: J. Montero, J. Stephansky, T. Cai, G.K. Griffin, L. Cabal-Hierro, E.A. Morgan, J.C. Aster, M.S. Davids, N.R. LeBoeuf, R.M. Stone, M. Konopleva, N. Pemmaraju, A. Letai, A.A. Lane

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Montero, J. Stephansky, T. Cai, K. Togami, E.A. Morgan

Study supervision: J. Montero, M. Konopleva, A. Letai, A.A. Lane

The authors thank Amanda Christie, Alexandra Christodoulou, David Weinstock, and the Dana-Farber Cancer Institute (DFCI) PDX repository (www.PRoXe.org) for helping with BPDCN xenografts; Takahiro Maeda (Nagasaki University) for sharing the CAL1 cell line; and Joel Leverson (AbbVie) for providing venetoclax, navitoclax, and A-1331852.

This work was supported by a DFCI Medical Oncology Research Grant (A. Letai and A.A. Lane), the Ludwig Cancer Research Foundation (J. Stephansky and J.C. Aster), and an American Society of Hematology Scholar Award (A.A. Lane).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.