The BCORL1 mutation frequency increased during the lineage switch process in patient P01.
The first patient (P01) was a 27-year-old male diagnosed with common B-ALL, exhibiting no atypical phenotypic features and fusion genes (Supplemental Table 1). Upon diagnosis, he was found to have mutations in FLT3-ITD and BCORL1 through the targeted sequencing (Supplemental Table 2). The chromosomal analysis indicated a karyotype of 46,XY,t(9;20)(p12;q12)[17]. After a brief remission post-chemotherapy, he experienced a relapse and was enrolled in a clinical trial for CD19-based CAR-T treatment. Notably, a lineage-switched leukemia was detected via flow cytometry analysis 5 weeks after CAR-T infusion. The analysis revealed 82.5% of cells expressing CD38, CD33, CD36, and HLA-DR robustly, along with the presence of CD13, CD123, and TDT. Moreover, there was partial expression of CD34, CD11b, CD64, CD11c, weak expression of CD15 and MPO, and an absence of expression for CD117, CD14, cCD79a, CD19, CD22, and CD20. This patient developed new STAG2 mutation in the bone marrow during the lineage switch while retaining the FLT3-ITD and BCORL1 mutations (Supplemental Table 2). Following that, he attained remission once more after undergoing a myeloid chemotherapy regimen and subsequently underwent hematopoietic stem cell transplantation. Currently, the patient continues to maintain a sustained remission.
In order to explore the potential mechanism of the lineage switch process in P01, we collected bone marrow or peripheral blood samples at four time points: before CD19 CAR-T therapy without bone marrow lymphodepletion (T1_Pre_CART), 28 days after CD19 CAR-T therapy (T2_CART_D28), after myeloid relapse (T3_Relapse), and after achieving remission through chemotherapy (T4_Post_Chemo). Single-cell targeted genomic sequencing and single-cell membrane protein sequencing were performed on these samples (Fig. 1A).
Through single-cell genomic sequencing and quality control, we obtained a total of 6,566 high-quality cells with related gene mutations (FLT3, BCORL1, and STAG2) at the four time points for clone structure inference (Supplementary Table 3). The Tapestri Insights analysis (Fig. 1B) revealed that the FLT3-ITD mutation consistently persisted at four different time points. Pre-existing BCORL1 mutation rapidly expanded after the initiation of myeloid relapse, while STAG2 mutation occurred with the presence of BCORL1 mutation. The clone (C1) only included FLT3-ITD mutation decreased at the T3_Relapse time point (67.2%, 65.1%, 2.7%). Instead, the clone (C4 and C2) included FLT3-ITD and BCORL1 mutations became predominant (16.1%, 4.3%, 96.1%). Simultaneously, compared with the T1_Pre_CART time point, we found that BCORL1 mutation burden (Fig. 1C) and copy number (Supplementary Fig. 1A) increased at the T3_Relapse time point. This suggested that the expansion of the BCORL1 mutation might be a key factor in myeloid relapse.
On the other hand, we obtained 27,517 high-quality cells with membrane protein information at two time points: T1_Pre_CART and T3_Relapse (Fig. 1D). At the T1_Pre_CART time point, B-ALL blast cells expressed typical B-ALL-associated markers CD19 and CD10, along with co-expression of myeloid-associated markers CD33 and CD123 (Supplementary Fig. 1B-C). However, after myeloid relapse, the blast cells lost the expression of lymphoid marker CD19, confirming the phenomenon of lineage switch. Both membrane protein sequence data and clinical flow cytometry data suggested the presence of myeloid lineage markers before myeloid relapse, supporting the hypothesis that a critical population with the potential for myeloid differentiation may exist in B-ALL blasts cells.
The CLP-like blast and pro-B-like blast exhibited the myeloid transformation potential in P01.
To further identify key populations with the potential for myeloid differentiation in B-ALL blasts in patient P01, we collected bone marrow samples at four time points for single-cell transcriptome sequencing (Fig. 2A) which were the same as time points selected for single cell targeted genomic and membrane protein sequencing.
Overall, 36,360 high-quality BMMNCs were obtained through rigorous quality control. Using 20 healthy donors as control, we generated a scRNA-seq reference map using 58,492 bone marrow cells (GSE120221, Supplementary Fig. 2A-B)23. The P01 bone marrow sample data were then projected to the reference data (Supplementary Fig. 2C-D). Unsupervised clustering analysis identified 12 clusters (Fig. 2B) with distinct maker genes (Supplementary Fig. 3A-C). Before CD19 CAR-T treatment, a subset of CD19-positive B-ALL cells, namely CLP-like blast and Pro-B-like blast, expressed myeloid lineage molecules like CD33 and CD123 (Fig. 2C), which were in accordance with membrane protein sequencing results. They also expressed lymphoid (EBF1) and myeloid (SPI1) transcription factors (Fig. 2D), suggesting that this subset had the potential for lymphoid and myeloid bi-directional differentiation. After CD19 CAR-T treatment, the CD19-positive B-ALL cell subpopulations were eliminated, leaving a residual subset expressing high levels of CD34, defined as the pre-GMP-like blast. The pre-GMP-like blast expressed CD33 and CD123, along with myeloid-related transcription factors (FOXO1, SPI1), potentially originating from the CLP-like blast. After myeloid relapse, the bone marrow was primarily composed of myeloid progenitor cells (GMP-like blast and myeloid-cell-like blast), and the expression of lymphoid-related transcription factors gradually disappeared (Supplementary Fig. 3D). Following chemotherapy, this population of myeloid progenitor cells was eliminated, and the bone marrow returned to a state of normal hematopoiesis.
To elucidate the hierarchy and developmental relationships among cell subsets, the pseudotime analysis was conducted using Monocle. The results (Fig. 2E) revealed that CLP-like blasts and myeloid blasts were positioned at the two ends of the transformation process, while the pre-GMP-like blasts were distributed along the intermediate differentiation path from B-ALL blast cells to myeloid blast cells. This implies that the pre-GMP-like blast, with the potential for myeloid differentiation as described above, may serve as a transitional subgroup during the myeloid switch program.
Besides, we observed the presence of BCR sequences (Fig. 2F) and FLT3-ITD mutation (Supplementary Fig. 3E) in blast cells before and after the myeloid relapse. This suggested that the origin of the myeloid progenitor cells could be traced back to B-ALL blast cells. In addition, unsupervised clustering of relevant subgroups based on highly variable genes (Fig. 2G) showed that this subgroup clustered with pre-GMP-like blast, indicating similarity in their transcriptomes. Pathway enrichment analysis of differentially expressed genes in this subgroup (Fig. 2H, Supplementary Table 4) revealed that the upregulated genes were enriched in AML, cell cycle, and p53-related tumor signaling pathways, while genes related to lymphoid differentiation were downregulated. This indicated that the pre-GMP-like blast did not exhibit lymphoid differentiation characteristics but rather demonstrated transcriptomic features of myeloid progenitor cells.
In summary, during the myeloid lineage reprogramming process in the P01 patient, we defined a pre-GMP-like blast as the transitional cluster. Additionally, FLT3-ITD and BCORL1 mutations, along with the expression of myeloid lineage flow cytometry markers, were observed in B-ALL blasts before the myeloid lineage transformation.
The BCOR mutation frequency increased at relapse in another lineage switch patient P02.
Patient P02, a 26-year-old male also diagnosed with common B-ALL (Supplementary Table 1), presented with the EP300::ZNF384 fusion gene and mutations in GNB1, and CCND3 at diagnosis (Supplementary Table 2). Following induction chemotherapy in line with our clinical trial protocols, this patient achieved complete remission. Subsequently, he underwent CD19 CAR-T therapy upon relapse following three courses of consolidation chemotherapy. Nevertheless, one-month post CAR-T therapy, he developed AML with a distinct phenotype from the ALL, as flow cytometry revealed 10.5% of cells with strong expression of CD34, CD13, CD33, and CD123, partial expression of CD36, weak expression of CD38, CD11b, CD4, and no expression of CD117, HLA-DR, MPO, TDT, CD14, and other myeloid and lymphoid markers. In addition, he was resistant to subsequent treatment attempts and died 3 months after CAR-T treatment.
To investigate the underlying factors contributing to the lineage switch in P02, we performed whole-exome sequencing and bulk BCR sequencing (Fig. 3A) on bone marrow cells obtained at two distinct time points: before CD19 CAR-T treatment (T1_Pre_CART) and 35 days after CD19 CAR-T treatment (T2_Relapse). Through whole-exome sequencing, we identified 31 and 33 non-silent somatic mutations at the T1_Pre_CART time point and the T2_Relapse time points, respectively (Supplementary Table 5). Additionally, integrating the targeted sequencing results from the time at initial diagnosis and minimal residual disease (MRD) time points, we generated a somatic mutation spectrum for P02 at different time points (Fig. 3B).
Moreover, according to the mutation spectrum, we inferred the dynamics of the clonal evolution structure of P02 at different stages of treatment (Fig. 3C). Throughout the treatment course of P02, the fusion gene EP300::ZNF384 persisted in expression, the IKZF2 mutated clone disappeared after CD19 CAR-T therapy, and the BCOR gene mutated clone emerged upon myeloid lineage relapse. Meanwhile, based on the specific mutational imprint, Signature.9 disappeared at the T2_Relapse time points (Fig. 3D).
Through BCR sequencing, we observed the presence of identical immunoglobulin sequences at the T2_Relapse time point as those in the T1_Pre_CART time point (Fig. 3E, Supplementary Table 6), and we could also observe that the clonal frequency of some immunoglobulin sequences increased during relapse, suggesting their enrichment in the myeloid reprogramming process. This result robustly proved that the origin of myeloid progenitor cells was reprogrammed from B-ALL cells. Therefore, like the findings in the P01 patient, we hypothesize that, in the case of the P02 patient, there might be a population with the myeloid-differential potential presented in B-ALL blasts.
The pro-B-like blasts exhibited the potential for bi-directional differentiation towards both lymphoid and myeloid lineages before CD19 CAR-T treatment.
To further ascertain whether there was a cluster in patient P02 with the potential for myeloid differentiation in B-ALL blasts, we conducted single-cell transcriptome sequencing on total bone marrow cells obtained before CD19 CAR-T treatment (T1_Pre_CART) and 35 days after CD19 CAR-T therapy (T2_Relapse) (Fig. 4A). After rigorous quality control, we obtained a total of 13,292 high-quality bone marrow cells. Through unsupervised clustering analysis, a total of 12 subgroups were identified (Fig. 4B) and annotated by projection to scRNA-seq data from healthy donor bone marrow samples23 (Supplementary Fig. 2). At the T1_Pre_CART time point, the bone marrow primarily consisted of CD19-positive B-ALL progenitor cells (Pro-B-like blast, Pre-B-like blast, and Immature B-like blast), immature B cells (corresponding to the clinical flow cytometry CD34+CD19+ B lymphoblast subpopulation), mature B cells, and plasma cells. At the T2_Relapse time point, in addition to normal immune cell subpopulations, there were also myeloid progenitor cells (GMP-like blast) in the bone marrow (Supplementary Fig. 4A-B).
Through the analysis of single-cell transcriptome data, we sought to identify cells carrying mutation information before and after myeloid transformation (Fig. 4C). We observed that the main gene mutated clone appearing in pro-B-like blasts also appeared in the myeloid blasts. This finding indicated myeloid blasts originated from the pro-B-like cluster.
In our quest to identify the cluster with the potential for myeloid differentiation, we employed Monocle trajectory inference on blast cells before and after myeloid relapse. The findings unexpectedly revealed two differentiation pathways (termed Path I and II) with distinct transcription factors (Fig. 4D-E). Different clusters merged into a process in pseudotime that started with the pro-B-like blast (mostly in the pre-branch) and ending with the subset GMP-like blast (Path I) or subset immature-B-like blast (Path II), respectively. Signature genes of different expression paradigms were identified (Fig. 4F), including genes expressed from Pre-branch to Path I (such as COX8A), from Pre-branch to Path II (such as BCL11A), and specifically expressed on Path I (such as S100A9).
The pathway enrichment analysis of characteristic genes in the pro-B-like blast subpopulation (Supplementary Fig. 4C, Supplementary Table 7) revealed upregulation in myeloid cell differentiation, B cell differentiation, oxidative phosphorylation, cell cycle, and epigenetic regulation pathways like the PRC2 pathway. Conversely, there was a downregulation of genes associated with hematopoietic cell lineage differentiation. These results suggest that the subpopulation undergoing myeloid reprogramming originates from an earlier stage, characterized by stronger proliferative capacity within the pro-B-like blast group.
In summary, we have defined a pro-B-like blast subpopulation with bi-directional differentiation potential towards the lymphoid and myeloid lineages and traced the accumulation of BCOR mutation in patient P02.
The role of BCOR and BCORL1 truncating mutations in B-ALL cohort.
By analyzing the two rare B-ALL lineage switch cases, we interestingly observed a substantial increase of truncating mutational burden in two homologous genes: the BCOR and BCORL1. These mutations resulted in the loss of the same functional domain, namely the PUFD/PUFD-like domain, preventing the recruitment for the formation of the PRC1.1 complex (Fig. 5A, Fig. 5C, Supplementary Fig. 5A).
To explore the role of BCOR and BCORL1 mutations in B-ALL, we conducted a retrospective analysis of targeted sequencing from 349 newly diagnosed B-ALL patients in our center. Among the 349 B-ALL patients, 32 cases (9.17%) carried a total of 48 BCOR/BCORL1 mutations (Fig. 5D, Fig. 5B). Simultaneously, these 32 patients harbored kinase related gene mutations such as NRAS, KRAS, and FLT3 (Fig. 5E), with NRAS and BCOR co-deletion playing a crucial role in leukemia progression24. Remarkably, one of the 32 patients with a BCOR truncating mutation transformed into B/myeloid mixed phenotype acute leukemia during treatment (Fig. 5F). P01 and P02 with myeloid transformation also carried truncating mutations, suggesting a potential impact of BCOR/BCORL1 truncating mutation on cellular plasticity. On the other hand, through retrospective analysis, we found that B-ALL patients carrying BCOR/BCORL1 truncating mutations exhibited higher myeloid markers like CD33 and CD123 at the initial diagnosis compared with B-ALL patients without BCOR/BCORL1 truncating mutation (Supplementary Fig. 5B).
In summary, through retrospective analysis, we have observed that patients carrying these two gene mutations exhibited differentiation multipotency, and the coexistence with RAS may play a crucial role in the occurrence of leukemia.