Blastic Plasmacytoid Dendritic Cell Neoplasm Are Not Characterized By Two Distinct Subgroups, but By a Transcriptomic and Phenotypic Gradient

Background: Blastic Plasmacytoid Dendritic Cell Neoplasm (BPDCN) is a rare haematological malignancy characterized by NF-κB activation (Sapienza et al., 2014), but also BCL2 overexpression (Montero et al., 2017), UV exposure damages (Griffin et al., 2023), neural signatures (Sapienza et al., 2021) and fatty acid metabolism (Ceroi et al., 2016). However, BPDCN is heterogenous: recent studies showed two subgroups, based on immune response involving CD11b/CD177 (Summerer et al., 2021) or on pure pDC/conventional DC-enriched signatures (Künster et al., 2022). Künster et al. also showed NF-κB signalling pathway versus EZH2 dependence (Künster et al., 2024).

Methods: We conducted a retrospective study on bone marrow (BM) aspirates and peripheral blood (PB) samples obtained at diagnosis (French pDC network ROMI DC-2008-713/DC-2016-27-91, Besançon, France). Our previously published transcriptomic data of 13 BPDCN (GSE89565) were used as a training cohort and a second cohort (n=37) was used as a validation to determine two subgroups of BPDCN, using bulk RNAsequencing (NEBNext Ultra II, New England Biolabs, Ipswich, MA, on NextSeq500, Illumina, San Diego, CA) on sorted populations (lymphocytes and blastic pDCs). Data were compared to 6 pDC and 4 lymphocyte populations obtained from healthy donors (HD) from French Blood Establishment. Paired MethylSequencing (MethylSeq) was performed on 30 blastic pDCs with SureSelectXT (Agilent, Santa Clara, CA) on NextSeq500 (Illumina). In-house pipelines on R were developed for transcription factor activities (DoRothEA/VIPER package), fusion transcript (Arriba/FusionCatcher/Starfusion), mutations (GATK), alternative splicing (Whippet), deconvolution (BisqueRNA), T-cell clone diversity (MIXCR), inference between blastic pDCs and lymphocytes (BulkSignalR), methylation status (Bismark/Methylkit/DMRichR and clusterProfiler). Probability of overall survival (OS) was determined using the Kaplan-Meyer method and log-rank test.

Results: Two clusters were identified in the training cohort (median age= 61y[15-82], 11 males/13), one cluster with only men exhibiting MYB/MYC rearrangements (n=6) and another with apparent shorter OS, however without significance (n=7). These two clusters were not entirely transposed in the validation cohort (median age=67y[8-93], 26 males/37). As expected, epigenetic/splicing factor genes were highly mutated and MYB fusions were detected. Two clusters may be defined on blast fraction, but they were not strongly distinguished by unsupervised clustering. Interestingly, the lymphocyte fraction signature was shown to reflect the blasts. Combining z-scores of the 2 fractions, we note a correlation with two extreme cluster E1 and E2 without two distinct groups, but a continuity. E2 highly expressed pDC markers, HLA-II, IFN-I signalling, DNA replication, UV damages and fatty acid β-oxidation pathways. Deconvolution showed a various microenvironment, with higher T-cell clone diversity (unique TCRA/B). In contrast, E1 showed cell-cell signalling, NF-kB pathway, neural, hematopoiesis and TLR4 signatures in blasts, associated with T-cell activation, Th2 profile, IL4/IL13 signalling and ILC2 signatures,. Moreover, exon skipping was enhanced in contrast to E2 and HD pDCs. The inference study between blasts and lymphocytes showed interactions involving IL18, EGFR, IL13 in E2 and CD14, FGF in E1. Methylseq confirmed an involvement of neural and Wnt pathways in E1, while E2 showed hypomethylation of cellular response and immunological synapses targets. At last, despite short survival, our group 1 exhibited an adverse prognosis (OS 4.3 months vs 9.4 months, p=0.0431).

Conclusions: even if BPDCN can be divided in two subgroups, this strategy splits intermediate patients who are very similar. We prefer to define a continuity between 2 extreme clusters. Despite the lack of a clear-cut distinction, the clusters harbour specific biological features and clinical impact. E1 confirm the C2 cluster of Künster et al., 2024 on an independent cohort. Our E2 cluster would be associated with UV damage, replication activation and fatty acid metabolism while E1 exhibit less differentiated signature, with NF-κB activation, neural signatures and different ability to chat with lymphocytes compared to E2. Prognosis and therapeutic impacts are hypothesized, NF-κB being targetable by proteasome inhibitors for instance.

Renosi:Servier: Other: congress registration; Incyte: Other: speaker for an event, participating to another. Deconinck:Menarini-Stemline: Other: consultancy, Research Funding; ImmunoGen: Research Funding; Abbvie: Research Funding; Novartis: Research Funding.