Blastic Plasmacytoid Dendritic Cell Neoplasm: State of the Art and Prospects

Blastic plasmacytoid dendritic cell neoplasm (BPDCN) is an extremely rare tumour, which usually affects elderly males and presents in the skin with frequent involvement of the bone-marrow, peripheral blood and lymph nodes. It has a dismal prognosis, with most patients dying within one year when treated by conventional chemotherapies. The diagnosis is challenging, since neoplastic cells can resemble lymphoblasts or small immunoblasts, and require the use of a large panel of antibodies, including those against CD4, CD56, CD123, CD303, TCL1, and TCF4. The morphologic and in part phenotypic ambiguity explains the uncertainties as to the histogenesis of the neoplasm that led to the use of various denominations. Recently, a series of molecular studies based on karyotyping, gene expression profiling, and next generation sequencing, have largely unveiled the pathobiology of the tumour and proposed the potentially beneficial use of new drugs. The latter include SL-401, anti-CD123 immunotherapies, venetoclax, BET-inhibitors, and demethylating agents. The epidemiologic, clinical, diagnostic, molecular, and therapeutic features of BPDCN are thoroughly revised in order to contribute to an up-to-date approach to this tumour that has remained an orphan disease for too long.


Definition
Blastic plasmacytoid dendritic cell neoplasm (BPDCN, ICD-O code 9727/3) is regarded as an orphan tumour due to its rareness and usual clinical aggressiveness with poor response to conventional chemotherapies [1]. It derives from precursors of plasmacytoid dendritic cells (pDCs), also known as professional type I interferon-producing cells or plasmacytoid monocytes. In the Revised WHO Classification of Tumours of Haematopoietic and Lymphoid Tissues, BPDCN is quoted after acute myeloid leukaemia [1]. This reflects the fact that the gene signature of the cell of origin is much closer to myeloid than lymphoid precursors [2]. reference to chronic myelomonocytic leukaemia, a fact that is not surprising taking into consideration the co-occurrence of several gene mutations (see below) [1].

Clinics
The disease tends to involve multiple sites [1]. More often, it affects the skin (in 60-100% cases), followed by the bone-marrow and peripheral blood (in 60-90% of cases) and lymph nodes (in 40-50% of cases). In the natural history of the disease, the skin is the first affected site [23][24][25][26][27][28] (90% of patients), where it usually remains confined until a rapid, second step (i.e., leukemic spread and multiorgan involvement) occurs, eventually leading to death. It has been hypothesised that the skin may act as a sanctuary organ limiting the disease spread at the beginning [29]. However, a few cases lacking the initial cutaneous involvement have been reported in the literature [30][31][32][33][34]. BPDCN cutaneous tropism has been related to the expression of skin-migration molecules such as CLA and CD56 by the neoplastic elements. Another possible explanation may be the local availability of chemokines binding cognate receptor expressed by the neoplastic cells such as CXCR3, CXCR4, CXCR6, CXCR7. At skin level, the disease can present as isolated or disseminated bruise-like lesions [27,35]. The lesions are usually described as erythematous to purplish papules, plaques or tumours with a heterogeneous size (from few millimetres to several centimetres) with no preferred anatomic area ( Figure 1) [27,35]. On clinical grounds, an important distinction should be made between the presence of isolated and eruptive lesions [29]. The former have a better clinical outcome, while the latter should be regarded as a marker of an aggressive disease (progression free survival of 23 vs. 9 months, respectively) [29]. Theoretically, the different behaviour may be due to a high tumour burden in the eruptive presentation. Cases featuring mucosal involvement, especially in the oral cavity have rarely been observed [28]. reference to chronic myelomonocytic leukaemia, a fact that is not surprising taking into consideration the co-occurrence of several gene mutations (see below) [1].

Clinics
The disease tends to involve multiple sites [1]. More often, it affects the skin (in 60-100% cases), followed by the bone-marrow and peripheral blood (in 60-90% of cases) and lymph nodes (in 40-50% of cases). In the natural history of the disease, the skin is the first affected site [23][24][25][26][27][28] (90% of patients), where it usually remains confined until a rapid, second step (i.e., leukemic spread and multiorgan involvement) occurs, eventually leading to death. It has been hypothesised that the skin may act as a sanctuary organ limiting the disease spread at the beginning [29]. However, a few cases lacking the initial cutaneous involvement have been reported in the literature [30][31][32][33][34]. BPDCN cutaneous tropism has been related to the expression of skin-migration molecules such as CLA and CD56 by the neoplastic elements. Another possible explanation may be the local availability of chemokines binding cognate receptor expressed by the neoplastic cells such as CXCR3, CXCR4, CXCR6, CXCR7. At skin level, the disease can present as isolated or disseminated bruise-like lesions [27,35]. The lesions are usually described as erythematous to purplish papules, plaques or tumours with a heterogeneous size (from few millimetres to several centimetres) with no preferred anatomic area ( Figure 1) [27,35]. On clinical grounds, an important distinction should be made between the presence of isolated and eruptive lesions [29]. The former have a better clinical outcome, while the latter should be regarded as a marker of an aggressive disease (progression free survival of 23 vs. 9 months, respectively) [29]. Theoretically, the different behaviour may be due to a high tumour burden in the eruptive presentation. Cases featuring mucosal involvement, especially in the oral cavity have rarely been observed [28].

Microscopic Findings
In its most common form, BPDCN is characterized by a diffuse, monomorphous infiltrate of medium-sized blasts reminiscent of either lymphoblasts or myeloblasts [1]. Nuclei have a slightly irregular profile, fine chromatin, and from one to several small nucleoli. The cytoplasm rim is usually narrow and turns greyish-blue and agranular on Giemsa staining. Mitoses are variable in number, and the Ki-67 rate ranges from 20 to 80% ( Figure 2). Recently, a morphologic variant provided with immunoblastic-like appearance has been reported in association with MYC rearrangement [36]. Angioinvasion and coagulative necrosis are absent [27,37]. In the skin, the dermis is usually massively infiltrated, with extension to the subcutaneous fat. The epidermis and adnexa are generally spared [27]. In lymph nodes, there is diffuse involvement of the interfollicular

Microscopic Findings
In its most common form, BPDCN is characterized by a diffuse, monomorphous infiltrate of medium-sized blasts reminiscent of either lymphoblasts or myeloblasts [1]. Nuclei have a slightly irregular profile, fine chromatin, and from one to several small nucleoli. The cytoplasm rim is usually narrow and turns greyish-blue and agranular on Giemsa staining. Mitoses are variable in number, and the Ki-67 rate ranges from 20 to 80% ( Figure 2). Recently, a morphologic variant provided with immunoblastic-like appearance has been reported in association with MYC rearrangement [36]. Angioinvasion and coagulative necrosis are absent [27,37]. In the skin, the dermis is usually massively infiltrated, with extension to the subcutaneous fat. The epidermis and adnexa are generally spared [27]. In lymph nodes, there is diffuse involvement of the interfollicular areas and medulla, B-cell follicles being more often spared. Bone marrow biopsy shows either a subtle interstitial infiltrate (detectable only by immunohistochemistry) or-more often-massive replacement; residual haematopoiesis may display dysplastic changes, especially in megakaryocytes [38]. On peripheral blood and bone-marrow smears, tumour cells may show cytoplasmic microvacuoles localized along the cell membrane and pseudopodia [39]. areas and medulla, B-cell follicles being more often spared. Bone marrow biopsy shows either a subtle interstitial infiltrate (detectable only by immunohistochemistry) or-more often-massive replacement; residual haematopoiesis may display dysplastic changes, especially in megakaryocytes [38]. On peripheral blood and bone-marrow smears, tumour cells may show cytoplasmic microvacuoles localized along the cell membrane and pseudopodia [39].
Although recurrent, none of these alterations turned out to be BPDCN-specific, also being observed in other hematological malignancies.
By FISH analysis, MYC translocations were reported in the 39% of BPDCN patients, in association with the above mentioned immunoblast-like morphology [36]. T(6;8)(p21;q24) corresponded to the commonest type of MYC rearrangement: it defined a subgroup of patients with a more aggressive behavior [69]. Of clinical relevance, MYC positivity was found to confer good response to the acute lymphoblastic leukemia (ALL)-based chemotherapy in a limited number of patients [36,70]. Furthermore, FISH analysis documented in a few cases the translocation of the MLL1 gene, also recorded in 18% of ALLs [71][72][73][74], along with frequent rearrangements of the ETS variant gene 6 (ETV6), a transcription factor disrupted in other hematological malignancies [75,76].

Gene Expression Profiling by Array
The first study of BPDCN gene expression profiling (GEP) was conducted in 2007 by Dijkman et al. Since BPDCN skin lesions could easily be confused with cutaneous myelomonocytic leukemia (c-AML), Dijkman et al. performed a-CGH and GEP by array of 5 BPDCN skin biopsies and 6 c-AML cases. According to their study, BPDCN displayed: (1) a transcriptome profile and a molecular karyotype indeed distinct from c-AMLs; (2) recurrent deletions of 4q34, 9, and 13q12-q31 chromosomal regions; (3) lower expression of RB1 and LATS2 tumor suppressor genes; (4) higher expression of various pDC-related genes, such as the TLRs, TLR9 and TLR10 [60]. In 2014, Sapienza et al. compared for the first time the gene signature of 27 BPDCN primary samples with that of normal pDCs and found that the tumor transcriptome was more similar to resting pDCs rather than activated ones, confirming at molecular level the origin of BPDCN from a pDC precursor [2]. Tumor samples displayed 142 differentially expressed genes, mostly upregulated (89%), including those encoding for CyclinD1 and the anti-apoptotic protein BCL2. Bioinformatic analysis of GEP data revealed the aberrant activation of the NF-kB pathway, a finding suggesting possible response of BPDCN samples/cell lines to the proteasome inhibitor Bortezomib [2]. In vitro and in vivo experiments demonstrated that Bortezomib successfully shuts-down the NF-kB pathway and significantly induces BPDCN cell apoptosis, providing a potential new therapeutic option for BPDCN patients [2,81].
Ceroi et al. performed transcriptional profiling of 12 BPDCN cases by array and focused on a specific signature of downregulated genes involved in cholesterol homeostasis and responsible for its accumulation within the tumor cells. These sets of downregulated genes, if activated, stimulated the cholesterol efflux from neoplastic cells, inhibited the NF-kB pathway and arrested the BPDCN tumor cell survival [82].

Sequencing Studies
The chromosomal lesions of BPDCNs fully reflect their myeloid origin and the same could be expected at the DNA mutational level. Starting from the premise that the mutations of the epigenetic regulator gene TET2 are diffused in the myeloid lineage [83], Jardin et al. decided to explore the mutational status of this gene in 13 BPDCNs. TET2 was mutated in more than half of patients and was mostly affected by deleterious mutations (frameshift or nonsense). At diagnosis, TET2 mutations (54%) were recurrently flanked by TP53 mutations (38%) leading to hypothesize a synergistic effect between the two genes [84]. Alayed et al. confirmed the high mutational frequency of TET2 in BPDCN [46]. Ladikou et al. conducted the first targeted-sequencing on the BPDCN circulating free DNA of BPDCN cases, by identifying novel mutations of TET2 and RHOA [85]. Besides TET2, thanks to the targeted sequencing approach, many other myeloid-associated genes have been investigated in BPDCN. Taylor et al. presented to the ASH Meeting the first study of targeted sequencing of 219 myeloid-related genes in seven BPDCN samples. The most frequently mutated gene was the splicing factor ZRSR2 (57%) ex aequo with TET2 (57%), followed by ASXL1, TP53 and IDH2, KRAS, ABL1, ARID1A, GNA13, U2AF1, SRSF2, and the transcription factor IRF8 associated with dendritic cell deficiency [14]. Overall, 50% and 20% of patients with mutations in genes encoding for epigenetic factors or belonging to the IKAROS family respectively experienced a significantly reduced overall survival [87].
More recently, integrated "omics" approaches have been applied aiming to better understand the tumor biology. Montero et al. analyzed, by RNA-sequencing, 12 BPDCN samples and four pDCs from healthy donors by confirming BCL2 overexpression in tumors. Furthermore, by the BH3-proling of two BPDCN cell lines (CAL-1 and GEN2.2), six primary patient samples, and six patient-derived xenografts, the same authors demonstrated the BCL2 dependence of BPDCN elements as well as their sensitivity to the BCL2 inhibitor venetoclax. In the light of this finding, two patients were then treated with venetoclax and experienced significant disease responses [88,89].
Ceribelli et al. first performed an RNA interference screening study of the CAL-1 BPDCN cell line and recognized the transcription factor TCF4 as a master regulator of the BPDCN oncogenic program: its downregulation provoked the loss of the BPDCN-specific gene expression signature along with tumor cell death. Already described as relevant in normal pDC development, the TCF4 gene product was positively detected by immunohistochemistry in all the 28 BPDCN samples examined and proposed as a new reliable diagnostic marker (see above) and potential therapeutic target for bromodomain and extra-terminal domain inhibitors (BETis) [13].
Emadali et al. further substantiated the use of BETis in BPDCN. They examined 47 tumor samples and the CAL-1 cell line by various techniques (e.g., cytogenetics, a-CGH, FISH, targeted sequencing) and found that the loss of the glucocorticoid receptor gene, NR3C1, defined a high-risk group of patients. NR3C1 is often juxtaposed with lncRNA3q, a novel nuclear noncoding RNA involved in the regulation of leukemia stem cell programs and G1/S transition and aberrantly overexpressed in BPDCN malignant cells. BETis successfully turned-off the expression of lncRNA3q and inhibited tumor cells growth [90].
Suzuki et al., used RNA sequencing technology to discover novel fusion genes in 14 BPDCNs corresponding to five children and nine adults; recurrent MYB gene rearrangement were identified in all the children (100%) and in four out of the nine adults (44%) [91]. Sapienza et al. analyzed BPDCNs by WES, RNA and Chromatin Immunoprecipitation (ChIP) sequencing approaches. Several epigenetic factor genes were found mutated (e.g., ASXL1, TET2, SUZ12, ARID1A, PHF2, CHD8) and the functional enrichment analysis of the mutational data showed that of all the biological programs explored, the epigenetic was the most affected. At transcriptomic level, the patients displayed the significant enrichment of gene signatures related to epigenetic pathways, predicting response to hypomethylating agents. Accordingly, the use of 5'-azacytidine in combination with decitabine significantly inhibited disease progression and extended survival in a preclinical mouse model [22].

Therapy of Blastic Plasmacytoid Dendritic Neoplasm
BPDCN is characterized by an inherent resistance to standard chemotherapies. Treatment responses are mostly transient, the overall outcome being general very poor in general [32,54,92]. Given the rarity of the disease, the available data on BPDCN therapy mainly derive from retrospective studies.
In general, intensive induction regimens (e.g., hyperCVAD) are considered more effective compared to standard therapies (e.g., CHOP-like) [39,92,93]. In general, ALL-like treatments seem to be more effective in term of response rates than AML-like induction therapies [32,54,92]. The inclusion of l-asparaginase in ALL-like regimens could be a significant determinant of efficacy in this setting, as l-asparaginase has shown clinical activity in BPDCN in combination with single agent methotrexate [94,95].
Regarding the role of hematopoietic stem cell transplant in BPDCN therapy, there are several reports suggesting better results in terms of enduring remissions and relapse rates with allogeneic-stem cell transplantation (allo-SCT) compared to auto-SCT. These studies demonstrated durable complete remissions with allo-SCT, with OS rates ranging from 40% at 10 years, to 58% at 3 years depending on the follow-up period [96,97]. In general, allo-SCT consolidation seems to yield the best results when performed in first complete remission (CR) [97][98][99], with OS rates reaching 74-82% at 3-4 years [97,99]. Reduced intensity conditioning seems to be equivalent to myeloablative regimens in terms of relapse rates [97]. However, these data should be interpreted with due caution given the possible biases arising from the retrospective nature of these studies (e.g., patient selection bias, absence of intention to treat analyses, small sample size).
Eligible patients should be considered for allo-SCT consolidation in first CR whenever feasible. It should be noted, however, that these patients represent the minority of BPDCN patients, as the disease normally affects elderly patients, with a median age of 68 years [32].
For elderly patients, lower intensity treatments can be explored. Lower intensity chemotherapy regimens demonstrated some efficacy in BPDCN, such as single agent pralatrexate, bendamustine, or gemcitabine/docetaxel combinations [100][101][102][103]. However, despite the promising results, these studies were performed on a small number of patients, and which ought to be validated in larger future studies.
A recent study by our group strongly supports the use of hypomethylating agents, demonstrating a significant enrichment in epigenetic modifiers mutations in the setting of BPDCN [22]. In line with our preclinical findings, two clinical reports have demonstrated activity of 5-azacitidine in BPDCN, although the responses were generally transient once again [104,105]. Combinatory approaches based on hypomethylating agents should be explored in the near future.

Novel Agents
Given the unsatisfactory results of low-intensity treatments, and the toxicity of intensive therapies and allo-SCT consolidation, there is strong rationale for the use of novel targeted agents for the treatment of BPDCN.
SL-401 is a novel recombinant protein including components of diphtheria toxin fused to interleukin-3. As mentioned in previous sections, CD123 is expressed on the surface of BPDCN cells. In a phase I study of SL-401 in BPDCN the overall response rate was 77% (with 55% CR) in the evaluable patient population (seven out of eleven patients were able to complete the planned treatment) [58]. A phase 2 study reported at the 2017 ASH meeting showed promising results with a 79% CR rate in first line and 31% CR rate in relapsed/refractory patients [106].
Phase I trials are ongoing with other immunotherapies targeting CD123, such as bispecific antibodies, immunoconjugates, and chimeric antigen receptor (CAR)-T-cells [107]. In fact, recent data show promising activity of anti-CD123 CAR T-cells in acute myeloid leukemia and preliminary experiences support the future implementation of anti-CD123 CAR-T cell therapy in the BPDCN setting [108]. The BCL-2 inhibitor venetoclax has shown high single agent activity in myeloid malignancies [109,110] and is currently under evaluation in combination with induction chemotherapy and hypomethylating agents.
Several recently published reports have described the activity of venetoclax in the setting of BPDCN [89,[110][111][112]. Venetoclax given as single agent or in combination with hypomethylating agents was able to induce meaningful clinical responses in relapsed/refractory patients. Further therapeutic options may be represented by bromodomain and extra-terminal domain inhibitors (BETis), which has been tested in preclinical studies [13,90].

Conclusions and Perspectives
Although the criteria for the diagnosis of BPDCN are well-defined [1], the knowledge of the pathobiology of the tumour is still based on a limited number of contributions, which reflect its exceptional occurrence. The epigenetic regulation, activation of the NF-kB pathway, and resistance to apoptosis seem to represent the main biological players, which should be taken into consideration in designing innovative therapeutic strategies. BPDCN is in fact characterized by intrinsic resistance to standard chemotherapies. In young patients, intensive ALL-like induction regimens followed by allo-SCT consolidation is considered the most effective treatment strategy, leading to durable responses in a fraction of cases. Elderly patients (who represent the majority of BPDCN patients) remain an unmet medical need. Recently, hypomethylating agents, anti CD123 directed immunotherapies and the BCL-2 inhibitor venetoclax showed promising single-agent clinical activity. These observations, together with emerging preclinical data provide the rationale for the prompt clinical testing of combinatory approaches with curative intent.