Single-cell heterogeneity and dynamic evolution of Ph-like acute lymphoblastic leukemia patient with novel TPR-PDGFRB fusion gene

Philadelphia chromosome-like acute lymphoblastic leukemia (Ph-like ALL) is an aggressive neoplasm of B lymphoblasts, with a gene expression profile similar to that of Ph⁺ ALL but lacking the BCR-ABL1 translocation, and 5-year overall survival does not exceed 20% [1, 2]. It has been reported that drug resistance and subsequent relapse induced by heterogeneous blasts remain the leading causes of death among Ph-like patients [3]. Single-cell RNA sequencing (scRNA-seq) for B-cell ALL (B-ALL) revealed the extensive remodeling of the immune microenvironment during the progression [4], pre-existing CD19 negative subclones [5], and the presence of an exhausted T cell subset with remarkable heterogeneity [6]. However, cellular heterogeneity and dynamics of Ph-like ALL at single-cell resolution remains unknown.

Previous study has been reported that patients with PDGFRB rearrangements account for ~ 8% of the Ph-like cases [7]. Our Ph-like ALL patient was a 55-year-old woman who harbored a novel TPR-PDGFRB in-frame fusion gene which consists of exons 1–46 of the TPR with exons 11–23 of the PDGFRB (Fig. 1A, and Additional file 1: Table S1, S2). Initially, this patient was diagnosed as ALL common-B based on morphology and immunophenotyping, and conventional chemotherapy regimen was started (Additional file 1: Table S3 and Fig. S1). However, she developed rapid relapse twice following short-lived remission (Fig. 1B). Previous studies have reported that PDGFRB fusion patients with Ph-like ALL were refractory to conventional therapy but amenable to tyrosine kinase inhibitor (TKI) [7]. Single agent of imatinib was unable to reduce the lymphoblasts within 2 weeks, and then further she developed into central nervous system leukemia (CNS-L) with 96.5% lymphoblasts (Fig. 1B). Due to the enhanced inhibitory of tyrosine kinase activity and the infiltration ability across the blood brain barrier, one course of dasatinib combined with DOCPL regimen was administered (Fig. 1B, Additional file 1: Table S3). Then, complete remission (CR) was achieved and maintained for 4 months (Fig. 1B). Ponatinib was given because of relapse and unfit for chemotherapy, unfortunately, she achieved no response and finally died after 14 months from initial diagnosis (Fig. 1B).

Meanwhile, we evaluated the expression similarity between this patient and Ph⁺ ALL patients based on two publicly available datasets [8, 9] (Additional file 1: Additional file methods). The patient exhibited higher correlation coefficients with Ph⁺ ALLs than with other ALL subtypes (Additional file 1: Fig. S2A), and blended into the Ph⁺ cluster under the hierarchical clustering analysis on the basis of the Ph-like ALL signature (Additional file 1: Fig. S2B), indicating the Ph-like expression feature. Fluorescence in-situ hybridization (FISH) confirmed the PDGFRB break-apart (Fig. 1C), reverse transcription-polymerase chain reaction (RT-PCR) and Sanger sequencing validated the junction sequences of the TPR-PDGFRB fusion transcript (Fig. 1D, Additional file 1: Fig. S2C). The presumed TPR-PDGFRB protein retained the transmembrane and tyrosine kinase domains from PDGFRB (Fig. 1E), and immunoblotting confirmed the presence of chimeric protein TPR-PDGFRB (Fig. 1F). Furthermore, we observed the reduced and elevated frequency of TPR-PDGFRB fusion transcripts along with disease remission and recurrence (Additional file 1: Fig. S2D-G), suggesting the potential as a marker for evaluating the minimal residue disease (MRD) of this patient.

To decipher the cell heterogeneity and dynamics of Ph-like ALL, scRNA-seq was performed on 10,273 cells taken from the bone marrow specimens at diagnosis and first relapse (Fig. 2A, Additional file 1: Fig. S3A). Seventeen cell clusters were respectively labeled as B cells, megakaryocytes-erythroid progenitors (MEPs), T cells, natural killer (NK) cells, classical and non-classical monocytes (Fig. 2B, C, Additional file 1: Fig. S3B-F). Further, twelve B cell clusters were classified into six subsets involved in CD34⁺ ProB cells, CD20 (MS4A1) high and IGLL1 low PreB cells, CD20 low and IGLL1 high PreB cells, IGKC low and IGLL1 high PreB cells, cycling PreB cells (MIK67 high expression), and memory B cells (Fig. 2D, E, Additional file 1: Fig. S3G). Pseudotime inference revealed the distinct bifurcated architecture of cell trajectory, which appearing to start principally from the CD34⁺ ProB and cycling PreB cells, then moved toward two IGLL1 high PreB cell subsets, following to CD20 high or memory B cell subsets, implying a divergence in the transcriptional state (Fig. 2F).

t-Distributed Stochastic Neighbor Embedding (tSNE) representation was divided on the basis of disease status (Fig. 2G), and all non-B cell clusters were enriched in relapse surprisingly (Additional file 1: Fig. S4A, B). Primary-enriched B cell clusters 4, 6, and 15 (> 80%), along with the low expression of CD19 and high expression of CD79A and CD10 (Fig. 2G, H), showed down-regulation in the MYC targets and oxidative phosphorylation pathways (Additional file 1: Fig. S4C). Differentially expressed genes (DEGs) and significantly enriched pathways for the Relapse-specific B-cell clusters 2 and 14 (> 90%), characterized by the absence of CD20 expression, were identified (Fig. 2I, J). B-ALL patients with the high relapsed B-cell feature (Additional file 1: Additional file methods) was found to be significantly associated with a poor prognosis (P = 0.0077; Additional file 1: Fig. S4D). Furthermore, the expression of ACSM3 and HRK were higher in relapsed B-ALL with CNS-L, indicating their potential contribution to brain infiltration in Ph-like patient (Fig. 2K, Additional file 1: Fig. S4E). Taken together, these findings uncover the single-cell transcriptome heterogeneity and dynamic evolution during disease progression.

Next, we investigated the similarities and differences between Ph-like and Ph⁺ ALL at the single-cell level [4]. A total of 25,559 cells were classified into 23 clusters which are assigned into six different cell types (Additional file 1: Fig. S5A-C). We found that Ph-like specific B-cells (cluster 1, 8) showed a higher stemness feature of comparing with other B cell clusters (Additional file 1: Fig. S5D-F). In addition, we observed that cluster 10 (CD8⁺ T cells) from diagnosis and relapse samples (Additional file 1: Fig. S5D) possessed both high NK (KLRB1 positive) and high cytotoxicity signature scores (Additional file 1: Fig. S5G-I), consisting with those in solid tumors [10]. CLEC2D, as the ligand of KLRB1 (CD161), is reported to be expressed in T/NK cells, germinal center (GC)-associated B cells, early plasmablasts and GC-derived lymphomas, but not in B-ALL cells [11], and protect target cells against NK cell-mediated killing as the inhibitory immune checkpoint in NK cells [12]. Our scRNA-seq data revealed that the expression level of CLEC2D in malignant B-ALL cells (clusters 1, 2, 3, 5, 6, 8, 13, 19 and 22), was also comparable with that of T cells (clusters 4, 9, 10 and 16) and NK cells (clusters 15 and 23) (Additional file 1: Fig. S5J), implying the potential for immune evasion.

In recent years, immune microenvironment of tumors has already become more important in cancer treatment. Previous scRNA-seq analysis has revealed the pivotal role of monocytes [4] and exhausted T cells [6] in non-Ph-like B-ALL. Our study not only demonstrated the remarkable heterogeneity of malignant B-cells, but also discovered the ectopic CD8⁺ T cells and NK cells which may contribute to immune escape. However, our work lacks a further point in time of TPR-PDGFRB positive Ph-like patient and more Ph-like cases for validation. Moreover, subsequent design experiments are required to better elucidate the mechanism of the abnormal CLEC2D expression to guide the development of relevant molecular targeting drugs.