Identification and targeting of treatment resistant progenitor populations in T-cell Acute Lymphoblastic Leukemia

Abstract

Refractoriness to initial chemotherapy and relapse after remission are the main obstacles to cure in T-cell Acute Lymphoblastic Leukemia (T-ALL). Biomarker guided risk stratification and targeted therapy have the potential to improve outcomes in high-risk T-ALL; however, cellular and genetic factors contributing to treatment resistance remain unknown. Previous bulk genomic studies in T-ALL have implicated tumor heterogeneity as an unexplored mechanism for treatment failure. To link tumor subpopulations with clinical outcome, we created an atlas of healthy pediatric hematopoiesis and applied single-cell multiomic (CITE-seq/snATAC-seq) analysis to a cohort of 40 cases of T-ALL treated on the Children's Oncology Group AALL0434 clinical trial. The cohort was carefully selected to capture the immunophenotypic diversity of T-ALL, with early T-cell precursor (ETP) and Near/Non-ETP subtypes represented, as well as enriched with both relapsed and treatment refractory cases. Integrated analyses of T-ALL blasts and normal T-cell precursors identified a bone-marrow progenitor-like (BMP-like) leukemia sub-population associated with treatment failure and poor overall survival. The single-cell-derived molecular signature of BMP-like blasts predicted poor outcome across multiple subtypes of T-ALL within two independent patient cohorts using bulk RNA-sequencing data from over 1300 patients. We defined the mutational landscape of BMP-like T-ALL, finding that NOTCH1 mutations additively drive T-ALL blasts away from the BMP-like state. We transcriptionally matched BMP-like blasts to early thymic seeding progenitors that have low NR3C1 expression and high stem cell gene expression, corresponding to a corticosteroid and conventional cytotoxic resistant phenotype we observed in ex vivo drug screening. To identify novel targets for BMP-like blasts, we performed in silico and in vitro drug screening against the BMP-like signature and prioritized BMP-like overexpressed cell-surface (CD44, ITGA4, LGALS1) and intracellular proteins (BCL-2, MCL-1, BTK, NF-κB) as candidates for precision targeted therapy. We established patient derived xenograft models of BMP-high and BMP-low leukemias, which revealed vulnerability of BMP-like blasts to apoptosis-inducing agents, TEC-kinase inhibitors, and proteasome inhibitors. Our study establishes the first multi-omic signatures for rapid risk-stratification and targeted treatment of high-risk T-ALL.

Figure 1. Arrest states of T-ALL subtypes and other acute leukemias in reference to human hematopoiesis.

(a) Selection of 25 ETP-ALL, 5 Near-ETP ALL, and 10 Non-ETP T-ALL patients from the Children’s Oncology Group AALL0434 cohort (n=1411) based on response to induction therapy (Day 29 MRD). (b)UMAP representation of bulk RNA-Seq data from 1335 diagnostic T-ALL samples from COG AALL0434. Each point represents the bulk-RNA-seq transcriptome for one patient. Patients selected for single cell study are indicated by circular points. All ETP patients are colored red. (c) UMAP representation of CITE-seq (n=271,603 cells, 271,603 plotted) and scATAC datasets n=332,663 cells; due to size of peak x cell matrix, 60,000 randomly downsampled cells are plotted), colored by level 1 annotations (detailed in Fig. S1). (d) UMAP representation of healthy human hematopoiesis development reference trajectories, based on scRNA (left) and scATAC (right) data. Key stages of T-cell development implicated are labeled. (e) Arrest state of leukemic cells from 40 T-ALL patients based on projection to healthy scRNA (left) and scATAC (right) reference. D value from two-sample Kolmogorov-Smirnov test is indicated to the side of brackets (*, p<2.2e-16). 10 T/M MPAL and 10 AML patients sequenced using identical assays (Chen et al., in preparation) are included as comparator groups. (f) Proportion of ETP-blasts in four key T-cell developmental stages, as compared to other T-ALL patients (left). Proportion of ETP-blasts in three key myeloid developmental stages, as compared to T/M MPAL and AML patients (right). P-values are based on two-sided Mann-Whitney test (*** < 0.001; ** < 0.01, * < 0.05). Results based on scRNA projection are shown. BMP stage encapsulates multipotent progenitors: HSPC/LMPP/CLP/ETP. α/β stage encapsulates all cells that have moved past T-commitment: DP/ α/β / α/β (mature) / Naïve T.

Figure 2. Treatment Resistance in ETP-ALL is associated with a progenitor-like population.

(a-b) Arrest state of leukemic cells from High MRD and MRD-Negative ETP-ALL patients based on projection to healthy scRNA (left) and scATAC (right) reference trajectory. A, Proportion ranges from 0–0.3. b, Proportion ranges from 0–0.5. D values from two-sample Kolmogorov–Smirnov test are indicated by the brackets (*, p< 2.2e-16). © Stratification of n=25 single-cell-sequenced ETP-ALL patients by BMP-like proportion (High: >30%; Low < 30%). (d) Differentially expressed genes between BMP-like blasts from non-responding patients and T-specified blasts from responding patients. © Stratification of n=113 ETP patients from AALL0434 using 119-gene DEGs between BMP-like and T-specified blasts obtained in panel d. Prognostic value of the BMP-like-DEG signature in multivariate analysis (with Day 29 MRD and CNS status) is shown below the Cox-proportional hazard log-likelihood p-value. Left: stratification with signature alone; Right: stratification with signature & EOI MRD status.

Figure 3. NOTCH1 mutations are associated with the T-specified rather than BMP-like state.

(a) recurrent driver fusions (n=16, recurrent in 2+ samples) among 113 bulk-sequenced ETP-ALL patients (Polonen et al) and associated BMP-like and T-specified signature scores. (b) Recurrently mutated genes (n=26, recurrent in >5 samples) among 113 bulk-sequenced ETP-ALL patients (Polonen et al) and associated tumor BMP-like and T-specified signature scores. (c) Arrest state of leukemic cells from NOTCH1 WT (n=19) and NOTCH1activated (n=6) leukemias based on scRNA and scATAC developmental trajectories. D value from two-sample Kolmogorov-Smirnov test is indicated to the side of brackets (*, p<2.2e-16). (d)Proportion of leukemic cells in BMP-like and T-lineage (Pro-T : αβ) cell states in NOTCH1 WT and NOTCH1 mutant in single-cell profiled patients. (e) T-specified signature score among 110 bulk-sequenced ETP-ALL patients from AALL0434. Patients are divided into three groups by NOTCH1 mutation status. P-values from two-sided Mann-Whitney test are shown (* < 0.05; ** < 0.01, ***< 0.001). (f) Overall survival of n=113 AALL0434 ETP patients by NOTCH mutation status. The p-value for log-likelihood statistic of Cox-proportional hazard test run with Day 29 MRD as a co-variate is shown in the bottom left.

Figure 4. A consensus, 17-gene BMP-like signature predicts overall survival across all subtypes of T-ALL.

(a) Arrest state of leukemic cells from CR and MRD+ Non-ETP-ALL patients based on projection to healthy scRNA (left) and scATAC (right) reference trajectory. (b) Arrest state of pre-committed Non-ETP blasts in CR and MRD+ patients. BMP-like encapsulates all cells that possess multipotent potential: HSPC/LMPP/CLP/ETP. (c)Overlap of ETP BMP-like and Non-ETP BMP-like DEGs to create consensus signature for risk stratification in AALL0434 (fully sequenced) and AALL1231 (partially sequenced). BMP-like DEGs were filtered for mean Log2FC > 0.9 between ETP and Non-ETP comparisons. (d) BMP-17 signature score within BM/thymus scRNA-seq reference. Multipotent BM progenitor populations with high BMP-17 AUC are circled. (e-i), Kaplan-Meier plot showing overall survival of bulk-RNA-sequenced T-ALL patients in AALL0434 and AALL1231 binarized using the BMP-like-17 signature. Prognostic value of the BMP-like-17 signature in multivariate analysis (with Day 29 MRD and CNS status) is shown below the Cox-proportional hazard log-likelihood p-value.

Figure 5. Nomination and pre-clinical validation of targeted therapy against BMP-like blasts.

(a) UMAP representation of 16 primary patient samples and corresponding PDX models profiled using scRNA-seq. PDX engrafted blasts are connected to their primary sample via arrows. (b) proportion of BM projected blasts (HSPC/LMPP/CLP) in 8 patient-PDX pairs. Samples are ordered by proportion of BM-projected blasts in the primary sample (left). Patients PATTDP, PAVYVY, PASZMC, PAVVVF had induction failure with BMP-like blasts > 25%. (c) Computational screening approach used to identify targetable genes within BMP-like blasts. 552 BMP-like blast specific DEGs (FDR < 0.05) were overlapped with 3 different drug target databases (TTD, DrugIDB, OpenTargets), LINCS1000 (a transcriptomic-based compound screening database), and DepMap (cancer cell line crispr-dependency database). Each gene was assigned a database score (+1 for each database) and differential expression score (ranging from 0–3). The top 10 targets by aggregate score are highlighted in red. (d) A panel of 40 drugs (left) was tested on PDX engrafted blasts from 5 T-ALL patients (2 high BMP-like and high MRD, 1 high BMP-like with relapse, 1 low BMP-like MRD- ETP, 1 low BMP-like MRD- Non-ETP). (e) Relative activity of drugs active in High BMP-like vs Low-BMP-like leukemias. (f) Dose response curves for nominated therapeutics that showed differential activity in high-BMP-like vs low-BMP-like leukemias. Response to prednisolone is shown as a comparison.