Deciphering the cellular hierarchy of multiple myeloma uncovers a therapeutically exploitable EP300/TBX2 axis in LILRB4+ myeloma-initiating cells

Abstract

Abstract Background: In contrast to leukemia, where the cancer stem cell model has clearly explained tumor hierarchies, multiple myeloma (MM) lacks a unified understanding of its cellular hierarchy, impeding both biological understanding and therapeutic advances. Early attempts to classify MM hierarchies relied on morphological criteria or surface markers. However, these studies failed to establish a developmental continuum linking MM subtypes to normal plasma cell maturation. The putative origin of MM hierarchies‒whether from myeloma stem cells (MSCs), myeloma-initiating cells (MICs), or clonotypic memory B cells‒remains contentious. The lack of consensus in defining myeloma developmental hierarchies—particularly their divergence from normal plasma cell maturation—has hindered the development of MIC-targeting therapies for MM. To address this gap and develop therapies against MICs, we reconstructed the MM cellular hierarchy based on normal plasma cell development, functionally identified putative MICs, and discovered a promising MIC-targeting signaling axis. Methods: We utilized both open-access and self-generated single-cell transcriptomic data of normal and MM plasma cells to construct a developmental blueprint. Computational analysis of stemness and developmental progression identified surface markers of myeloma populations resting at earlier developmental stage, which were then sorted by flow cytometry for functional validation. An epigenetic drug library was screened to identify compounds capable of depleting primitive myeloma populations. CUT&Tag was employed to investigate molecular mechanisms. Results: First, we established an eight-stage developmental framework for normal plasma cells and computationally mapped single-cell transcriptomes from MM specimens, thereby constructing a comprehensive MM hierarchy blueprint. Based on this framework, deconvolution analysis of bulk RNA-seq data of MM patients categorized them into six distinct subgroups according to their distribution pattern of the eight developmental stages. The six subgroups correlated with cytogenetic profiles and presented significant prognostic values. Further, we identified two stages scored highest in stemness including: 1) cycling plasma cells in the peripheral blood and bone marrow (PC-cyclingPB/BM), 2) low-CXCR4 plasma cells in the PB/BM (PC-lowC4PB/BM), as did their corresponding PC-cyclingPB/BM -like and PC-lowC4PB/BM -like myeloma cell counterparts. Surface marker profiling identified LILRB4 as a consistent marker of the two high-stemness stages in both normal and malignant plasma cells. Patients with higher LILRB4 expression showed poorer prognosis. Relapsed patients demonstrated increased LILRB4+ cells compared to newly diagnosed cases. We next validated the functional output of LILRB4+ cells. Single LILRB4+ MM cells generated both LILRB4+ and LILRB4-colonies, whereas LILRB4- cells produced only LILRB4-ones. Furthermore, LILRB4+ MM cells exhibited greater colony-forming unit (CFU) capacity and established stronger tumor burden in vivo. Following bortezomib (BTZ) treatment, LILRB4+ MM cells demonstrated higher viability than their LILRB4- counterparts. These suggest LILRB4 as a putative marker of MICs resistant to BTZ. Finally, we utilized an epigenetic drug library to identify compounds capable of inducing differentiation in LILRB4+ MM cells. EP300 inhibitors emerged as the top candidates that downregulated LILRB4 expression in these cells, simultaneously abolishing BTZ resistance. Animal studies evaluating EP300 inhibitors as a key differentiation-inducing therapy in combination with BTZ are currently in development. Mechanistically, CUT&Tag-seq revealed that EP300 does not directly bind to the LILRB4 promoter, but binds to transcription factor TBX2, highly expressed in LILRB4+ cells. Consistently, EP300 inhibition resulted in downregulation of TBX2. TBX2-knockdown reduced LILRB4 expression, subsequently decreased survival following BTZ treatment, reduced CFU potential, and diminished tumorigenic capacity in vivo. These changes phenocopied the effects observed with EP300 inhibitors. Conclusions: Our study provides the first comprehensive blueprint of MM hierarchy, identifying functionally validated LILRB4+ MICs residing at the apex of the developmental hierarchy as potential drivers of disease initiation and relapse, and nominates EP300 inhibition as a promising differentiation-based therapeutic strategy through TBX2 downregulation.