Mutation and cell state cooperation drives progression and is a targetable feature of remission in acute lymphoblastic leukemia

Abstract

Methods for treating leukemia are disclosed based on detecting specific cell states and transcriptional programs within leukemic cells. This disclosure presents a novel therapeutic method for treating acute lymphoblastic leukemia (ALL), including BCR-ABL positive and BCR-ABL1-like ALL subtypes. The method involves detecting specific cell states and transcriptional programs in patient samples and administering targeted therapies based on these characteristics. For a pre-B cell-like state or pre-BCR signaling program, a combination of tyrosine kinase inhibitor (TKI) and SYK inhibitor is used. Conversely, a progenitor-like state or stress-autophagy program is treated with a TKI and a p38 MAPK inhibitor. This approach aims to improve treatment efficacy by tailoring therapy to the leukemia's unique molecular and cellular features, particularly in relapsed cases or when minimal residual disease is present.

Description

MUTATION AND CELL STATE COOPERATION DRIVES PROGRESSION AND IS A TARGETABLE FEATURE OF REMISSION IN ACUTE LYMPHOBLASTIC LEUKEMIACROSS REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Application No. 63 / 675,115 filed July 24, 2024. The entire contents of the above-identified applications are hereby fully incorporated herein by reference.STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH

[0002] This invention was made with government support under Grant No.(s) CA217377, and CA23319501 awarded by the National Institutes of Health. The government has certain rights in the invention.REFERENCE TO AN ELECTRONIC SEQUENCE LISTING

[0003] Reference is made to the electronic sequence listing ("BROD-5965WP_ST26.xml"; Size is 4,621 bytes, created on July 24, 2025) is herein incorporated by reference in its entirety.TECHNICAL FIELD

[0004] The application is designed for the treatment of leukemia. Compositions and methods are disclosed for targeting and monitoring transcriptional and mutational states in residual leukemia cells that contribute to therapeutic resistance, relapse, and disease progression.BACKGROUND

[0005] Approximately 1.6 million Americans are diagnosed with cancer every year, and approximately 580,000 people in the United States are expected to die of the disease in 2013. Over the past few decades, cancer detection, diagnosis, and treatment have undergone significant improvements, resulting in increased survival rates. Nevertheless, despite the ever- increasing availability of new targeted therapeutics, only about 60% of people diagnosed with cancer are still alive five years after the onset of treatment, which makes cancer the second leading cause of death in the United States.

[0006] Cancers are primarily an environmental disease, with 90-95% of cases attributed to environmental factors and only 5-10% to genetics. Environmental factors encompass any cause that is not inherited genetically. Common environmental factors that may contribute to cancer death include tobacco (25-30%), diet and obesity (30-35%), infections (15-20%), radiation (both ionizing and non-ionizing, up to 10%), stress, lack of physical activity, and environmental pollutants.

[0007] Each cancer is characterized by the site, nature, and clinical cause of undifferentiated cellular proliferation, whereby the underlying mechanism for cancer initiation is unknown.

[0008] Cancer is traditionally treated with chemotherapy, radiation therapy, and surgery. Chemotherapy, in addition to surgery, has proven helpful in several different cancer types, including breast cancer, colorectal cancer, pancreatic cancer, osteogenic sarcoma, testicular cancer, ovarian cancer, and certain lung cancers. Radiation therapy involves the use of ionizing radiation in an attempt to either cure or improve the symptoms caused by cancer. It is used in approximately half of all cases, and radiation can be from either internal sources, in the form of brachytherapy, or external sources. Radiation is typically used in addition to surgery and chemotherapy. Still, it may be used alone for the treatment of certain types of cancer, such as early head and neck cancer. For painful bone metastasis, radiation can alleviate pain in about 70% of people. The effectiveness of radiation and chemotherapy is often limited by toxicity to other tissues in the body. Anticancer therapies are frequently rendered ineffective because of the resistance of the tumor cells to radio- and chemotherapy.

[0009] Blood cancer generally includes three main types: lymphoma, leukemia, and myeloma. Lymphoma refers to cancers that originate in the lymphatic system. Lymphoma includes, but is not limited to, Hodgkin's lymphoma, non-Hodgkin's lymphoma (NHL), diffuse large B-cell lymphoma (DLBCL), and peripheral T-cell lymphomas (PTCL), etc. Leukemia refers to malignant neoplasms of the blood-forming tissues. Acute leukemia involves predominantly undifferentiated cell populations, whereas chronic leukemia involves more mature cell forms. Acute leukemia can be further divided into acute lymphoblastic leukemia (ALL) and acute myeloblastic leukemia (AML) types. The Merck Manual, 946-949 (17th ed. 1999). Chronic leukemia can be divided into chronic lymphocytic leukemia (CLL) or chronic myelocytic leukemia (CIVIL). The Merck Manual, 949-952 (17th ed. 1999). Myeloma is a type of cancer that affects plasma cells in thebone marrow. Because myeloma frequently occurs at many sites in the bone marrow, it is often referred to as multiple myeloma (MM).

[0010] Despite ongoing progress, all too often, first-line therapy treatment of cancer patients leads to an initial remission that eventually relapses as a result of progressive resistance to the therapy. Novel methods of identifying and preventing resistance to treatment are therefore urgently needed.

[0011] Citation or identification of any document in this application is not an admission that such a document is available as prior art to the present disclosure.SUMMARY

[0012] According to an embodiment, the present disclosure provides compositions and associated methods for treating, ameliorating at least one symptom or indication, or inhibiting cancer growth in a subject. In further embodiments, methods are provided for delaying the development of a tumor or preventing tumor recurrence.

[0013] The present disclosure outlines a comprehensive method for treating lymphoproliferative diseases, e.g., leukemia, based on detecting cell state and transcriptional programs within the cancer cells. The disclosed methodology is directed towards administering a tailored combination of therapeutic agents determined by the unique cellular characteristics and signaling pathways active in the cancer cells derived from a subject.

[0014] In an embodiment, the method includes identifying the cell state of leukemic cells obtained from a subject. The cell state is categorized as either a pre-B cell receptor (Pre-BCR) state or a stress / autophagy cell state. A specialized treatment regimen is prescribed depending on the detected cellular state and associated transcriptional program. This regimen involves administering a tyrosine kinase inhibitor (TKI) in combination with either a SYK inhibitor when a pre-B cell-like state or a Pre-BCR signaling program is detected or a p38 inhibitor when a progenitor-like cell state or stress-autophagy program is identified.

[0015] In an embodiment, the disclosure treats oncogene-addicted leukemias, including acute lymphoblastic leukemia (ALL), with particular emphasis on BCR-ABL1 B-cell acute lymphoblastic leukemia (B-ALL). These conditions are characterized by distinct genetic and molecular profiles, requiring innovative therapeutic strategies, especially when the leukemia is in relapse following an induction therapy that includes a TKI.

[0016] Functionality and biophysical attributes of leukemic cells, such as cell mass, can be measured to determine the cell state. A higher cell mass can indicate a progenitor-like state, while a lower cell mass can identify a pre-B cell-like state. Specific cell mass ranges are provided to distinguish between these two states accurately.

[0017] The treatment may be initiated at the beginning of treatment or administered after the initial treatment during the minimal residual disease (MRD) phase in subjects who exhibit residual leukemia cells after initial treatment. Additionally, a gene expression score can be calculated for the subject's leukemic cell sample to further classify the cells based on their cell cycle status and propensity towards senescence.

[0018] Therapeutic agents employed in the disclosed method may include, but are not limited to, TKIs, such as imatinib, dasatinib, nilotinib, bosutinib, and ponatinib; and SYK inhibitors, including, for example, fostamatinib, entospletinib, TAK-659, R406, cerdulatinib, and BAY 61-3606; and p38 inhibitors like losmapimod, ruxolitinib, SB 203580, doramapimod (BIRB 796), and panapimod (R-1503).

[0019] Furthermore, the disclosed method encompasses strategies to combat disease progression, particularly in subjects who have achieved remission yet present with MRD. The method is adaptable for various types of B-lymphoproliferative disorders and is particularly suited for oncogene-addicted leukemia forms, such as ALL and BCR-ABL1 B-ALL. First-line therapies may include catalytic and allosteric inhibitors of BCR-ABL1 tyrosine kinase and secondary mutations, which might drive relapse.

[0020] In an embodiment, the disclosed method may reduce the disease burden and sustain remission in subjects with leukemia.

[0021] In an embodiment, methods, and treatment regimens offer a potential new paradigm in the personalized treatment of leukemia.

[0022] In an embodiment, the cancer or tumor is a heme cell tumor or malignancy. In an embodiment, the cancer or tumor is a B-cell tumor. In an embodiment, the B-cell tumor can be, for example, Hodgkin's lymphoma, non-Hodgkin's lymphoma, follicular lymphoma, small lymphocytic lymphoma, lymphoplasmacytic lymphoma, marginal zone lymphoma, mantle cell lymphoma, diffuse large B-cell lymphoma, B-cell lymphomas, lymphomatoid granulomatosis, Burkitt's lymphoma, acute lymphoblastic leukemia, hairy cell leukemia, or B cell chronic lymphocytic leukemia.

[0023] In an embodiment, the present disclosure provides compositions and associated methods and kits to detect, treat, or prevent leukemia, such as acute lymphoblastic leukemia (ALL).

[0024] In an embodiment, a method of treating leukemia is disclosed comprising detecting in a sample from a subject suffering from leukemia a cell state and transcriptional program, where the cell state is a pre-B cell-like cell state or a progenitor-like cell state, and where the transcriptional program is a pre-B cell receptor (Pre-BCR) program or a stress-autophagy program, and where if the pre-B cell-like cell state and pre-BCR signaling program is detected, then administering to the subject a treatment comprising a tyrosine kinase inhibitor (TKI) and a SYK inhibitor, including, for example, fostamatinib, entospletinib, TAK-659, R406, cerdulatinib, and BAY 61-3606, and where if the progenitor-like cell state and stressautophagy program is detected, then administering to the subject a TKI inhibitor, including, but are not limited to, imatinib, dasatinib, nilotinib, bosutinib, and ponatinib, and a p38 inhibitor, for example, losmapimod, ruxolitinib, SB 203580, doramapimod (BIRB 796), or panapimod (R-1503).

[0025] In an embodiment, the leukemia can be oncogene-addicted.

[0026] In an embodiment, the leukemia can be acute lymphoblastic leukemia.

[0027] In an embodiment, the oncogene-addicted leukemia can be BCR-ABL1 B-cell acute lymphoblastic leukemia (B-ALL).

[0028] In an embodiment, the subject can be suffering a relapse of leukemia following induction therapy.

[0029] In an embodiment, the induction therapy includes treatment with a TKI.

[0030] The cell state can be a functional measurement of one or more phenotypes in an embodiment.

[0031] The cell state can be detected in an embodiment using a biophysical measurement.

[0032] In an embodiment, the biophysical measurement can be cell mass.

[0033] In an embodiment, a higher cell mass indicates the progenitor-like cell state and a lower cell mass indicates the pre-B cell-like cell state.

[0034] In an embodiment, the higher cell mass can be between 15 and >25pg, and the lower cell mass can be between 10 and 15pg.

[0035] In an embodiment, the biophysical measurement can be size, mass, morphology, mechanical properties, or cellular stiffness.

[0036] In an embodiment, the TKI can be, for example, imatinib, dasatinib, asciminib, nilotinib, bosutinib, or ponatinib.

[0037] In an embodiment, the SYK inhibitor can be, for example, fostamatinib, entospletinib, TAK-659, R406, cerdulatinib, or BAY 61-3606.

[0038] In an embodiment, the p38 inhibitor can be, for example, losmapimod, ruxolitinib, SB 203580, doramapimod (BIRB 796), or panapimod (R-1503).

[0039] In an embodiment, the treatment is administered to a subject with minimal residual disease (MRD).

[0040] In an embodiment, the sample from the subject comprises minimal residual disease cells.

[0041] In an embodiment, the method includes determining a cycling and senescence gene expression score for the sample from the subject.

[0042] In an embodiment, the cycling and senescence gene expression score for the subject's sample can be categorized into one of three categories: (A) cells scoring as in cell cycle; (B) cells poised to enter cell cycle; and (C) cells scoring for senescence that have likely terminally exited the cell cycle.

[0043] In an embodiment, administering the treatment to the subject reduces disease burden and cancer remission.

[0044] In an embodiment, the transcriptional program includes a Pre-B cell receptor (PreBCR) program or a stress / autophagy program. In an embodiment the Pre-BCR program comprises one or more genes selected from the group consisting of; IGLL1, VPREB3, TCL1 A, UHRF1, H1F0, SOX11, RPS4Y1, VDAC1, CCND3, ARPP21, HMGN2, RPS26, CMTM8, IGJ, AOX2P, TMSB4X, FAM129C, SNURF, PTMA, LAT2, HIST1H3G, TP53INP1, SNRPN, NELLI, DNMT1, HMGB1, CBX1, PHGDH, LGALS9, EBF1, FLU, RPS4Y2, C16orf54, STMN1, ERGIC1, IRF2Bp2, PAXS, E2F2, CD38, GNG7, POU2AF1, UBE2T, MME, NUSAP1, DOK3, PVRIG, PIP4K2A, UCP2, and DEK. In an embodiment the stress / autophagy program comprises one or more genes selected from the group consisting of; DNAJB1, HSPA1A, HSP90AB1, HSPE1, HSPH1, DNAJA1, HSPA1B, NR4A1, HSPA8, CDKN1A, HBEGF, HSP90AA1, HSPB1, HSPA6, HSPD1, ZFAND2A, TUBB4B, EIF5, UBC, EIF1, DNAJB6, DDX3X, CHORDCI, JUN, ZC3HAV1, IER3, HERPUD1, IFRD1, BAG3, SAP18, HSA1, PMAIP1, BRD2, HSPA7, MAP1LC3B, SLC2A3, ZFAS1, TCP1,EIF4G2, ATF3, INSIGI, S0D2, SAT1, DEDD2, C6orf62, CACYBP, HEXIM 1, EGR3, and CSRNP1.

[0045] In an embodiment, a method is disclosed for extending progression-free survival in a subject receiving first-line therapy for a blood cancer comprising detecting, in a sample of a subject’s blood cancer cells, one or more secondary mutations driving relapse to said first-line therapy, measuring a gene expression profile characteristic of the cells’ developmental state, and administering a combination therapy comprising the first-line treatment in combination with a second-line therapy that targets a transcriptional pathway of the cells’ developmental state.

[0046] Extending progression-free survival (PFS) in cancer includes lengthening the period during which a patient's cancer does not worsen or metastasize following treatment.

[0047] In an embodiment, the subject can be in remission with minimal residual disease (MRD).

[0048] In an embodiment, the first-line therapy can be a catalytic and allosteric inhibitor of the BCR-ABL1 tyrosine kinase.

[0049] In an embodiment, the catalytic inhibitor may comprise ponatinib, and the allosteric inhibitor may comprise ascimininib (ABL001).

[0050] In an embodiment, at least one of the secondary mutations driving relapse can be a mutation in the BCR-ABL tyrosine kinase gene.

[0051] In an embodiment, one or more secondary mutations driving relapse may comprise a T3151 BCR-ABL mutation.

[0052] In an embodiment, the one or more secondary mutations driving relapse may comprise at least one other BCR-ABL mutation chosen from Y253H, F31 IL, and F359V.

[0053] In an embodiment, the one or more secondary mutations driving relapse may further comprise an activating mutation in STAT5A.

[0054] In an embodiment, one or more secondary mutations driving relapse may be a mutation in a gene encoding an ERK signaling pathway member.

[0055] In an embodiment, the gene encoding an ERK signaling pathway member may comprise KRAS, NRAS, BRAF, or PTPN11.

[0056] In an embodiment, if the first-line therapy comprises treatment with the catalytic inhibitor ponatinib, one or more secondary mutations driving relapse can activate mutations in the STAT5A gene.

[0057] In an embodiment, if the first-line therapy treatment with the allosteric inhibitor comprises ascimininib (ABL001), one or more secondary mutations driving relapse can activate mutations in a gene encoding an ERK signaling pathway member.

[0058] In an embodiment, if the first-line therapy comprises treatment with the allosteric inhibitor, ascimininib (ABL001) and the catalytic inhibitor ponatinib, one or more secondary mutations driving relapse can be activating mutations in a gene encoding an ERK signaling pathway member and in the STAT5A gene.

[0059] In an embodiment, a method of determining the developmental state of a cell sample taken from a subject with leukemia is disclosed, comprising isolating leukemic cells from the subject and measuring the buoyant cell mass at single-cell resolution, wherein leukemic cells with lower cell mass have a developmental state characteristic of a pre-B-like signature.

[0060] In an embodiment, a personalized treatment for a subject with cancer, including determining the genotypic expression in tumor tissue from the subject with cancer, includes assigning a gene expression score to the tumor, measuring the phenotypic cell state of cells from the tumor tissue from the subject with cancer to establish biophysical parameters describing the cell state; and treating the subject by administering an agent to treat the tumor tissue based on the gene expression score and cell state, and where the agent also stimulates the subject’s preexisting immune response.

[0061] In an embodiment, the determining the gene expression score includes sequencing polynucleotides from the tumor.

[0062] In an embodiment, the biophysical parameters include size, mass, morphology, mechanical properties, or cellular stiffness.

[0063] In an embodiment, the agent comprises a tyrosine kinase inhibitor (TKI) and a SYK inhibitor or a TKI inhibitor and a p38 inhibitor.

[0064] In an embodiment, the method further comprises predicting tyrosine kinase inhibitor sensitivity in BCR-ABL1 B-cell acute lymphoblastic leukemia by classifying leukemic cells according to B-cell developmental stage, wherein leukemic cells classified as mature B-cells demonstrate reduced tyrosine kinase inhibitor sensitivity compared to leukemic cells classified as progenitor B-cells.

[0065] These and other aspects, objects, features, and advantages of the example embodiments will become apparent to those with ordinary skill in the art upon considering the following detailed description of example embodiments.BRIEF DESCRIPTION OF THE DRAWINGS

[0066] An understanding of the features and advantages of the present disclosure will be obtained by reference to the following detailed description that sets forth illustrative embodiments, in which the principles of the disclosure may be utilized, and the accompanying drawings of which:

[0067] FIG. 1A-1D - Genetic mechanisms of resistance to oncogene withdrawal in BCR- ABL1 B-ALL. FIG. 1A - Phase Il-like randomized in vivo trial design, characteristics of the 13 PDX models used in the study, and mice profiled for genomic, immunophenotyping, or scRNA-seq at pre-treatment and progression time points. FIG. IB - Overall survival across the treatment arms in the Phase Il-like study; p-values are indicated for each pairwise comparison. FIG. 1C - Average variant allele frequency (VAF) for mutations along the ERK (y-axis) or STAT5 (x-axis) pathways averaged across mice in each PDX line at pretreatment or progression. Tied points represent time points from the same PDX line; black outlined points represent PDX lines derived from tumors with prior TKI exposure. Inset shows a subset of PDX models where Applicants did not detect a mutation in either pathway (n = 4). FIG. ID - Change in average VAF at progression for mutations along the ERK or STAT5 pathways compared to paired pretreatment tumors. Error bars represent + / - 1 standard deviation from the mean VAF.

[0068] FIG. 2A-2D -In vivo PDX trial outcomes. FIG. 2A - Serial peripheral blood qRT- PCR measurements from PDX model DF AB-13601 were treated with Ponatinib daily; each line represented an individual mouse. FIG. 2B - The competing risks model compares progression and non-progression mortality in mice by treatment arm; p-values indicate differences in progression and non-progression outcomes between treatment arms. FIG. 2C - Key trial events and outcomes for each mouse on trial; mice are grouped by treatment arm and annotated by pre-treatment risk score. FIG. 2D - Hazard ratios comparing pre-clinical risk factors for progression-free survival in PDX mice. Significant shifts (p<0.05) are annotated in red.

[0069] FIG. 3A-3B - Emergent patterns in BCR-ABL1 B-ALL mutation acquisition on TKI. FIG. 3A - Mutation profiles of individual PDX mice, grouped by treatment stage and annotated by treatment arm; summary of grouped pathway mutations included below. PDX models are annotated to assess patient treatment history and pre-treatment risk score. FIG. 3B - Change in the fraction of mice on each treatment arm harboring mutations between progression versus pretreatment time points. Genes along the STAT5 and ERK pathways are annotated in blue and red, respectively.

[0070] FIG. 4A-4H - Hybrid developmental transcriptional states define B-ALL. FIG. 4A - Simpson’s Diversity Index (SDI) of flow immunophenotyped B cell progenitor-like populations within individual tumors at pretreatment and progression; median SDI indicated for each group. Tied points represent pretreatment and progression samples from the same PDX line. ** indicates p<0.01. FIG. 4B - Fraction of PDX tumor at progression of each immunophenotyped B cell progenitor-like population. Tumors are grouped by mutation status at progression; bars represent the average tumor fraction, with error bars for + / - 1 standard deviation. Significant p-values from Dirichlet regression were noted: **p<0.01 and ***p<0.001. FIG. 4C - Schematic for deconvolving leukemic development-like transcriptional states using a random forest classifier trained on healthy bone marrow scRNA- seq data. FIG. 4D - K-nearest neighbor (kNN) projection for all leukemia cells onto reference normal hierarchy. FIG. 4E - Leukemic cells plotted according to non-Pro-B RF classification probabilities reveal hybrid developmental populations (Methods). Leukemic Pro-B cells lean towards HSC, Pre- or Immature B phenotypes. Red indicates cycling cells. FIG. 4F - Developmental marker gene co-expression in normal Pro-B cells (left) vs. leukemic cells classified as Pro-B cells (right). The X-axis represents the gene expression score difference between HSC (undifferentiated) and Pre-B and Immature B (more differentiated); the y-axis represents each cell’s second highest cell type marker expression score. Random samples of single cells from each bin are shown below. P-values compare expression distribution in normal vs leukemic Pro B cells for each normal cell type marker expression score (rows). FIG. 4G - Average expression of the top 30 genes correlated to RF prediction scores for each normal B-lineage cell type within each hybrid subpopulation (Methods). FIG. 4H - Scaled in silico predicted transcription factor (TF) activity over genes mutually informed with developmental hybrid gene signatures. Scaled TF activity scores shown in human reference samples and pre-treatment PDX lines, subset to TFs whose predicted activity scale with HSC, Pre-B, and Immature B classification confidence in leukemic cells.

[0071] FIG. 5A-5B - PDX tumor immunophenotyping. FIG. 5A - Flow gating strategy for B cell progenitor populations on a representative healthy human umbilical cord sample, PDX pre-treatment tumor, and PDX progression tumor (representative PDX=DF AB-62208). FIG. 5B - Fraction of PDX tumor across immunophenotyped B cell progenitor-like populations. Individual tumor fractions are plotted as points and annotated by pretreatment or progression time points; bars represent the average tumor fraction within each immunophenotyped population grouped by mutation, including error bars for + / - 1 standard deviation. Surface markers used for flow gating of each population, as shown in (FIG. 5A), annotated below.

[0072] FIG. 6A-6F - Generation of scRNA-seq healthy human reference of B cell development in the bone marrow. FIG. 6A - 8 healthy human bone marrow aspirates were flow sorted into live bulk, CFU-L (progenitor), Pre-Pro B, Pro-B, Pre-Pre-B, Pre-B, and Immature B populations for scRNA-seq profiling. FIG. 6B - Proportion of each scRNA-seq defined cell type from the bulk fraction (gray) or a flow sorted-fraction (green), emphasizing enrichment of low-abundance B cell progenitor populations in the reference dataset. FIG. 6C - Force-directed graph (FDG) projection of healthy human bone marrow annotated by hematopoietic cell types (n=13,643 cells). FIG. 6D - Dot plot of hematopoietic cell type marker genes. Color denotes scaled average expression; size denotes percent expression in each scRNA-seq cell type population. FIG. 6E - FDG projection of healthy human bone marrow, annotated by donor. FIG. 6F - Donor fractional contribution to each cell type population.

[0073] FIG. 7A-7F - Random Forest Classifier accurately classifies healthy and BCR- ABL1 B-ALL single-cell transcriptomes. FIG. 7A - 10-fold cross-validation of each healthy reference cell type during training. FIG. 7B - Top 200 random forest (RF) features ranked by permuted importances, grouped by healthy reference cell type (randomly downsampled n=100 cells). FIG. 7C - ROC curves for RF classification of test scRNA-seq bone marrow dataset; AUC listed in the inset for each cell type. FIG. 7D - Distribution of leukemic single-cell RF classification probabilities for each cell type, ordered by median RF classification probability. FIG. 7E - Pearson correlation of top 2000 highly-variable genes from healthy reference dataset across healthy and malignant cell type subpopulations. FIG. 7F - Scatter plots of each leukemic cell’s Shannon Diversity Index (SDI) of classification probabilities versus several UMI, genes,and percent mitochondrial transcripts. Cells removed from analysis due to the highest non-B cell lineage classification are outlined in red and colored by misclassified cell type.

[0074] FIG. 8A-8B - Defining B-ALL developmental tumor hybrid populations. FIG. 8A - Pearson correlation of in silico transcription factor (TF) activities across top mutually informed non-Pro B hybrid-defining TFs. FIG. 8B - Leukemic single cells were grouped into HSC-Like, Pro-B Hybrid, Pre B-Like, and Immature B-like malignant subpopulations according to their relative HSC-like, Pro B, Pre B-Like, and Immature B-Like gene expression module scores. Random Forest prediction probabilities, cycling or quiescent status, and PDX line are annotated for each cell.

[0075] FIG. 9A-9G - Oncogene withdrawal drives convergence onto developmental hybrids. FIG. 9A - Differentially expressed genes between PDX pre-treatment and progression single-cells. Marker genes for HSC, Pro B, Pre B, and Immature B cell types annotated. FIG. 9B - Hybrid population projections for all PDX leukemic cells, annotated by pretreatment or progression timepoints. Density distributions for pretreatment and progression time points along the x-axis and y-axis. FIG. 9C - Density of cells across the spectrum of hybrid developmental space, calculated by the difference between later-stage hybrid scores (Pre B- like, Immature B-like) and progenitor hybrid scores (HSC-like). Rows are annotated by PDX line, time point, and mutation status at progression. FIG. 9D - Simpson’s Diversity Index (SDI) of non-Pro B hybrid population proportions in each PDX line, colored by mutation status at progression. Tied points represented paired PDX time points. The P-value indicates a significant decrease in SDI between pretreatment and progression time points, excluding STAT5 pathway mutated mice (outlier DFAB-25157). FIG. 9E - Leukemic cells scored by the difference in hybrid mutually informed transcription factor activity across pretreatment and progression leukemic cells for three representative PDX lines. Density distributions for pretreatment and progression time points included across x- and y-axes. FIG. 9F -BCR-ABL1 qPCR traces from bone marrow aspirates of two patients on combination ABL1 inhibition, including one representative responder and one non-responder, from a Phase I clinical trial. FIG. 9G - Density of cells across the spectrum of hybrid developmental space, as defined in (FIG. 9D), compared across paired patient pre-treatment and on-treatment time point bone marrow aspirates. On the right, average module score for each sample showing increased or decreased ERK transcriptional activity over time.

[0076] FIG. 10A-10D - FIG. 10A - Hybrid population distributions for each profiled pretreatment and progression PDX mouse, annotated by treatment arm and time on treatment. FIG. 10B - Pre-treatment and progression average immunophenotyped population distributions for three representative PDX lines; error bars for + / - 1 standard deviation included and number of mice at each time point annotated. FIG. 10C - Scaled in silico predicted transcription factor (TF) activity over genes mutually informed with developmental hybrid gene signatures. Scaled TF activity scores are shown in human reference samples and in PDX- matched pre-treatment and progression time points, subset to TFs (Figure 41). FIG. 10D - Distribution of ERK signaling module scores within patient responder and non-responder leukemic cells at each pre-treatment and on-treatment time point. **p-adj<lE-5.

[0077] FIG. 11A-11G - Developmental phenotype is epistatic to mutations in remission. FIG. HA - Schematic for profiling three representative PDX models from pretreatment, remission, and progression with Smart-Seq2. FIG. 11B - t-SNE visualizations for the cells collected and labeled by the model (left), developmental state (middle), and genetic alterations (right). FIG. 11C - Density distributions of tumor cells at pretreatment, remission, and progression time points across progenitor-like to Pre B-like gene expression scores. FIG. HD - Single cells sorted by progenitor-like to Pre B-like gene expression scores within each time point and annotated by mutant or wild-type (WT) transcript detection for KRAS, NRAS, and PTPN11. Significant mutant transcript abundance between time points is annotated; *p<0.05 by Fischer exact test. FIG. HE - Dynamics of CNV subcl onal proportions at pre-treatment, remission, and progression in DFAB-25157; annotated by a fraction of each CNV subclone with detected KRAS or NRAS mutations. FIG. HF - Differentially expressed genes between DFAB-25157 KRAS or NRAS mutant cells in remission versus all other KRAS or NRAS mutant cells, highlighting increased expression of genes implicated in senescence. FIG. 11G - RAS- family mutant leukemic cells plotted according to their differentiation gene expression score on the x-axis and senescence gene expression score on the y-axis. Overlaid healthy progenitor, Pre-B, and Immature B cells are colored by cell type.

[0078] FIG 12A-12F - Random Forest Classifier recovers developmental structure in Smart-Seq2 single-cell transcriptomes. FIG. 12A - Proportion of random forest (RF) cell type classifications across all Smart-Seq2 (SS2) healthy and leukemic cells. FIG. 12B - Singlecells ordered by RF prediction probabilities from progenitor cell types to differentiated B cell types and annotated by flow sort gate. Below is the scaled expression of the top 10 RFprediction-correlated genes in developmentally-ordered healthy cells. FIG. 12C - Pearson cross-correlation of RF cell type prediction probabilities in SS2 healthy and malignant singlecells. FIG. 12D - Genes correlated to Pre-B RF prediction (x-axis), and genes correlated to Progenitor RF prediction are negatively correlated; rho and p-value from Pearson correlation noted. Colored points represent the top 30 progenitor and Pre-B correlated genes used to define the SS2 developmental spectrum. FIG. 12E - Leukemic SS2 single-cells ranked by Progenitor score, annotated by B cell-lineage RF prediction probabilities. Below is the scaled expression of the top 30 Progenitor-like and Pre B-like signature genes. FIG. 12F - Pearson crosscorrelation of RF cell type-correlated gene signature scores derived from Seq-Well and SS2 show cross-modality concordance.

[0079] FIG 13A-13F - SS2 enables co-detection of mutations and transcriptome in leukemic single-cells. FIG. 13A - Summary of recurrently identified ERK-family and STAT5- family mutations from bulk targeted sequencing across PDX lines profiled with Smart-Seq2. FIG. 13B - For genes with recurrently identified mutations, correlation of average gene expression and normalized mutation-locus detection rate (either mutant or wild-type reads). FIG. 13C - Mutant and wild-type transcripts detected in SS2 single-cell transcriptomes from three representative PDX mice; mutant transcript frequency in single-cells matched bulk variant allele frequency. FIG. 13D - SS2 single-cells within each profiled PDX line ordered by Progenitor-like scores,, Cycling status, CNV detection, and detected mutant and wild-type transcripts are annotated. FIG. 13E - t-SNE projection of SS2 single-cells from representative PDX lines CBAB-12402 (left), DF AB-62208 (middle), and DFAB-25157 (right), annotated by treatment time point. FIG. 13F - All RAS-family mutant leukemic single-cells grouped by three treatment time points, annotated by KRAS or NRAS mutant transcript detection, and ordered by progenitor-like signature score within DFAB-25157 and non-DFAB-25157 PDX lines.

[0080] FIG. 14A-14G - Biophysical measurements can be a surrogate for complex transcriptional states. FIG. 14A - Can complex gene expression states be represented by integrated biophysical features? FIG. 14B - Schematic for testing the hypothesis in (FIG. 14A) using sorted normal bone marrow populations and linked single-cell biophysical and transcriptional measurements (Methods). FIG. 14C - Mass distributions from the sorted populations in (FIG. 14B) were measured using the SMR. FIG. 14D - Leukemic cell mass- correlated genes on the x-axis versus the difference between leukemic genes correlated withrandom forest (RF) progenitor and Pre B cell types on the y-axis. Colored points mark genes included in the Progenitor and Pre B SS2 signatures; overall mass-development Pearson correlation and mass-developmental signature score Pearson correlations noted. FIG. 14E - Average RF prediction score (x-axis) versus average mass (y-axis) plotted for each mouse and annotated by progression mutation status. Pearson correlation demonstrates a positive correlation between progenitor transcriptomic phenotype and mass. FIG. 14F - Proposed workflow for comparing sequencing to biophysical measurements for diagnostics. FIG. 14G - Example application for pairing mutation and mass information to predict development and fitness-integrated transcriptomic state. Density spectra of (left) developmental score (Progenitor-like signature score minus Pre B-like signature score) and (right) mass for (top) healthy progenitor cells and immature B cells, and (bottom) RAS-mutant leukemic cells in three representative PDX lines. The dotted line represents (xx). Individual cells are colored according to their senescence signature score.

[0081] FIG. 15A-15C - Mass correlates with developmental states and cell cycle. FIG. 15A - Mass of healthy reference Smart-Seq2 (SS2) cells, binned by random forest-classified cell type and annotated by cell-type marker gene expression. FIG. 15B - Force-directed graph (FDG) visualization of healthy SS2 cells, annotated by cell type (top) and by cell mass (bottom); dot size indicates cell mass. FIG. 15C - Mass-correlated genes in healthy SS2 cells on the x-axis versus the difference between genes correlated with random forest (RF) progenitor and Pre-B cell types in healthy SS2 cells on the y-axis. Colored points denote marker genes for each cell type. R and p-value denote Pearson correlation between the x- and y-axis indicated gene correlations.

[0082] FIG. 16A-16H - Targeting integrative cell states enhances remission. FIG. 16A - Fitness landscape of cell-cycle arrested, poised, and active leukemic cells in remission. Singlecells plotted by cycling (x-axis) and senescence (y-axis) signature scores; quiescent-high cells annotated in black. FIG. 16B - Pairwise correlation of genes defining remission states, n = 30 genes in each program. FIG. 16C - Module scores for the stress-autophagy (turquoise) and pre-BCR signaling (dark red) states projected over single-cells at remission, plotted by their cycling (x-axis) and senescence (y-axis) gene signature scores. FIG. 16D - Study design for testing remission cell-state targeting. FIG. 16E and FIG. 16G - Cell fitness enrichment for DFAB25157 (top), and DFAB62208 (bottom). FIG. 16F and FIG. 16H- Remission diseaseburden assessed by flow cytometry (y-axis, relative to Ponatinib+ABLOOl) in the respective treatment arms.

[0083] FIG. 17A-17G - Targeting integrative cell states enhances remission. FIG. 17A - Single-cells at remission plotted by cycling (x-axis) and senescence (y-axis) signature scores. Quiescent-high cells are annotated in black, and RAS-family mutant cells are outlined in red. FIG. 17B - Single-cells at progression plotted as in (FIG. 17A). FIG. 17C - Single cells from DFAB25157 CNV subclone two at remission and progression ordered by fitness state and annotated by time point. FIG. 17D - All remission single-cells are ordered by pre-BCR signaling state scores, which are anti -correlated to stress-autophagy state scores. CNV and SNV mutation status annotated for each cell, along with cycling and quiescent status. FIG. 17E - Correlation between senescence signature score and progenitor-like gene expression versus cycling and Pre-B-like gene expression. FIG. 17F - Pathway enrichment p-values for the top 100 genes in the stress-autophagy remission state. FIG. 17G - Boxplot of distribution of relative MRD program (Pre-BCR signaling - Stress / Autophagy) in residual cells from DFAB-25157 and DF AB-62208.

[0084] FIG. 18A-18D - Ph+ cell lines with mature developmental states show increased resistance to ABL inhibition. FIG. 18A - UMAP visualization of three Ph+ cell lines (Z-181, Z-119, SUP-B15) demonstrating distinct transcriptional profiles. FIG. 18B - Dose-response curves showing differential sensitivity to dasatinib (left) and ponatinib (right) across the three cell lines, with IC50 values indicating varying drug sensitivity profiles. FIG. 18C - Random forest classification probabilities for each cell line across B-cell developmental stages, revealing progressively more mature B-cell signatures from Z-181 to Z-l 19 to SUP-B15. FIG. 18D - Correlation analysis demonstrating that cell lines with higher fractions of mature B-cell populations exhibit reduced sensitivity (higher IC50 values) to ABL inhibitors, validating the relationship between developmental cell state and therapeutic response.

[0085] FIG. 19A-19E - Genetic mechanisms of resistance to oncogene inhibition in Ph+ ALL. FIG. 19A - Motivation for evaluating efficacy and mechanisms of resistance to combination TKI therapy in Ph+ ALL. FIG. 19B - Patient characteristics of the 13 PDX models used in the study and Phase Il-like randomized in vivo trial design. Number of mice examined by genetic profiling, immunophenotyping ("Flow"), or scRNA-seq at pre-treatment and progression time points. For characteristics of patients from whom PDX lines were derived (Table S2): "TKI"=prior patient exposure to tyrosine kinase inhibitor; "relapse' -patient tumorat progression; "mut"=mutant (non-BCR: :AB / J) "p210" and "pl90"=p210 and pl90 BCR::ABL1 isoforms, respectively. FIG. 19C - Overall survival across treatment arms in Phase Il-like study; p-values from Cox regression analysis at clinical end-point (day 120) are indicated for each pairwise comparison between treatment arms. FIG. 19D - ABL and RAS pathway detected alterations in Phase Il-like study tumors at progression (n=40). Treatment emergent mutations indicated when mice from the same PDX line were profiled at pretreatment (see full alteration details in Figure 26A). Prior treatment indicates mice whose PDX lines were derived from patients with prior TKI and chemotherapy exposure. "VAF"=variant allele frequency. FIG. 19E - Average VAF for mutations along RAS (y-axis) or ABL (x-axis) pathways, averaged across mice in each PDX line at pretreatment or progression. Arrows link pretreatment and progression average VAFs from the same PDX line. PDX lines derived from patients with prior TKI exposure are outlined in black. Inset highlights a subset of PDX model timepoints where no (n=4 pretreatment, n=l progression) or few mutations were detected in either pathway.

[0086] FIG. 20A-20H - Hybrid developmental transcriptional states define B-ALL. FIG. 20A - Overview of Ph+ ALL scRNA-seq data collected from PDX lines (n=26,987 cells across 11 PDX lines inclusive of 38 pretreatment and progression tumors) and patient biopsies (n=l 5,680 cells across 5 patients inclusive of 14 pretreatment and on-treatment tumors). FIG. 20B - Unbiased factorization of leukemic scRNA-seq data with consensus non-negative matrix factorization (cNMF). Each row and column is an individual GEP and clustering is based on cosine similarity to find meta-programs (mGEPs; see Methods). "HSC"=hematopoietic stem cell; "ImmB ' -Immature B. FIG. 20C - Each mGEP annotated by the top 30 genes with the highest median cNMF gene spectra score across clustered intratumoral GEPs (Table S4). FIG. 20D - Approach for supervised classification using a random forest (RF) classifier trained on healthy bone marrow (BM) scRNA-seq data. FIG. 20E - Distribution (box plot and violin plot) of leukemia single-cell RF classification probabilities for each healthy BM cell type, ordered by median RF classification probability. "pDC"=plasmacytoid dendritic cell; "Ery"=erythroid; "Plasma' -plasma cell; "GMP"=granulocyte-monocyte progenitor; "Mono' -monocyte. FIG. 20F - K-nearest neighbor (kNN) projection of all leukemia cells onto reference normal hierarchy, annotated by number of classified leukemic cells for each reference B cell lineage population. FIG. 20G - Developmental marker gene co-expression in normal Pro-B cells (left) vs. leukemia cells classified as Pro-B cells (right). X-axis represents gene expression scoredifference between healthy HSC differentially expressed genes (undifferentiated) and the union of Pre-B and Immature B differentially expressed genes (more differentiated); y-axis represents each cell's second highest healthy cell type marker expression score. 300 randomly- sampled single cells from each bin are shown below. P-values from ANOVA (** / ?<0.001) compare expression distribution in normal vs leukemic Pro-B cells for each normal cell type marker expression score (rows). FIG. 20H - Leukemia cells plotted according to non Pro-B RF classification probabilities. Cells are colored by RF Pro-B classification probability (greyscale, fill) and cells are outlined by their classified cell type.

[0087] FIG. 21A-21H - Oncogene withdrawal drives convergence onto developmental hybrids. FIG. 21A - Simpson's Diversity Index (SDI) of non ProB-like hybrid population proportions in each PDX line, colored by mutation status at progression. Tied points represent paired PDX treatment stages. Median SDI for pretreatment and progression across PDX lines plotted as a line. Wilcoxon rank sum p-value (** / ?<0.01) reported, excluding ABL pathway mutated PDX line (outlier DFAB-25157). FIG. 21B - Differentially expressed genes between PDX pretreatment and progression single-cells. Marker genes for HSC, Pro-B, Pre-B, and Immature B cell types are annotated. FIG. 21C - Density of cells across the spectrum of hybrid developmental gene expression space, calculated by the difference between later-stage hybrid scores (PreB-hyb, ImmatureB-hyb) and progenitor hybrid scores (HSC-hyb). Rows are annotated by PDX line, time point, and mutation ("mut.") status at progression. FIG. 21D - SDI of flow cytometry immunophenotyped B cell lineage populations within individual PDX tumors at pretreatment and progression; median SDI indicated for pretreatment and progression tumors. **p<0.01 (Wilcoxon rank sum test). FIG. 21E - Fractional representation of immunophenotyped B cell lineage populations for 42 leukemia samples from 11 PDX lines at pretreatment and progression time points. "Pre." = pretreatment; "Prog." = progression. Immunophenotyped population flow cytometry markers defined in Figures S7B & S7C. FIG. 21F - Pretreatment and progression average immunophenotyped population proportions (as plotted in (E)) for three representative PDX lines corroborate transcriptional trends in (C); error bars indicate ±1 standard deviation when at least 3 mice were profiled. Number of mice profiled at each time point indicated for each PDX line. PDX lines are labeled based on mutation group at progression. FIG. 21G - BCR::ABL1 percent mRNA qRT-PCR traces (loglO(BCR::ABL / p-Actin mRNA)) from bone marrow aspirates of two patients on combination ABL1 inhibition, including one representative responder (BIAB-16768) and onenon-responder (DF AB-71417), from a Phase I clinical trial (Table S6). MRD 3.0 indicates trial definition of remission tumor burden (3-log reduction in bone marrow BCR::ABL1 mRNA detected by qRT-PCR). Right: scRNA-seq data collected from patients at each treatment cycle time point shown on t-SNE projections. FIG. 21H - Density of cells across the spectrum of hybrid developmental space, as defined in (C), compared across paired patient pre-treatment and on-treatment time point bone marrow aspirates.

[0088] FIG. 22A-22I - Developmental phenotypes restrict genotype fitness in remission. FIG. 22 A - Strategy for profiling three representative PDX models at pretreatment, MRD, and progression with Smart-Seq2 (SS2). FIG. 22B - t-SNE visualizations for the leukemic cells collected with SS2 and labeled by PDX line (top), developmental state (middle; "dev. ' -development), and detected genetic alterations (bottom; "SNV"=single nucleotide variant; "CNV"=copy number variant). FIG. 22C - Density distributions of leukemia cells at pretreatment, MRD, and progression time points across HSC-hyb to PreB-hyb gene expression scores. * / ?<0.001 from KS test for each pairwise comparison between treatment stages. FIG. 22D - Mutant or wild-type (WT) transcript detection for KRAS, NRAS, and PTPN11 within single-cells. Significant mutant transcript abundance between time points are annotated; * / ?<0.05 by Fisher exact test. FIG. 22E - Dynamics of CNV sub-clonal proportions at pretreatment, MRD, and progression in DFAB-25157. Pie charts represent KRAS or NRAS fraction of each sub-clone at the indicated time points. Number of cells sampled within each CNV sub-clone are reported. FIG. 22F - Differentially expressed genes between DFAB-25157 AV S'-mutant cells at MRD versus all other AV S'-mutant cells, highlighting increased expression of genes implicated in senescence (Table S8). FIG. 22G - RAS-pathway mutant leukemic cells plotted according to their differentiation gene expression score on the x-axis, and senescence-like gene expression score on the y-axis. Overlaid healthy progenitor, Pre-B, and Immature B cells colored by cell type. FIG. 22H - Fitness landscape of cell-cycle arrested, poised, and actively cycling leukemic cells in remission. Single-cells plotted by cycling (x- axis) and senescence (y-axis) signature scores. Distributions for cells in each fitness quadrant shown (green=MRD, red=Progression; * / ?<0.001 reported from KS test). FIG. 221 - DFAB- 25157 leukemic cells from each CNV subclone ranked along senescence-like and cell cycle signature scores. Fisher's exact test p-value reported for the origin of cycling cells (MRD vs. Progression). No cells belonged to the "poised" fitness category from either CNV subclone.

[0089] FIG. 23A-23G - Targeting integrative cell states enhances remission. FIG. 23A - Pairwise Pearson correlation of genes defining MRD states (Table S9). FIG. 23B - Module scores for the Stress- Autophagy (turquoise) and Pre-BCR Signaling (dark red) states projected over single-cells at MRD. Cells are plotted along fitness quadrants as in Figure 22H by their cycling (x-axis) and senescence-like (y-axis) gene signature scores. FIG. 23C - Study design for testing MRD cell-state targeting. FIG. 23D - Cell fitness distribution for DFAB-25157 MRD cells. FIG. 23E - DFAB-25157 MRD bone marrow disease burden assessed by flow cytometry (y-axis, relative to Ponatinib+Asciminib) in the respective treatment arms ("Asc."=Asciminib; "Fos. ' -Fostamatinib; "Los."=Losmapimod). T-test p-values reported, comparing losmapimod and fostamatinib arms to asciminib reference. FIG. 23F - Cell fitness distribution for DF AB-62208 MRD cells. FIG. 23G - DF AB-62208 MRD disease burden assessed by flow cytometry as in (E). Reported t-test p-values compare losmapimod and fostamatinib arms to asciminib reference.

[0090] FIG. 24A-24G - Biophysical measurements can be used as a surrogate for complex transcriptional states. FIG. 24A - Schematic for evaluating the relationship between complex transcriptional state and integrative biophysical features. FIG. 24B - Mass distributions from the sorted populations in (A) measured using the SMR; median mass reported. FIG. 24C - Leukemia cell mass-correlated genes (x-axis) are plotted against each gene's correlation to developmental phenotypes (RF probability for progenitor and Pre-B cell types; y-axis). Colored points mark genes included in the Progenitor and Pre-B SS2 signatures; "Sig. genes' -Leukemia developmental marker genes. FIG. 24D - Average difference in RF prediction score between early and late stages of B cell development (x-axis) versus average mass for each mouse (n=17), binned by distributions in (B) and Figure 35A, and annotated by progression mutation status. FIG. 24E - Proposed workflow for comparing sequencing to biophysical measurements for diagnostics. FIG. 24F - Example application for pairing mutation and mass information to predict development and fitness-integrated transcriptomic state. Density spectra of (left) developmental score and (right) mass for (top) healthy progenitor cells and immature B cells, and (bottom) RAS-mutant leukemic cells in three representative PDX lines. Dotted line for mass distribution represents mean+1 standard deviation of healthy Immature B mass. Median differentiation scores or mass for each PDX line are denoted as a dot; PDX lines are colored based on their median similarity to Immature B or Progenitor differentiation scores or mass. * indicates significant difference between DFAB-25157differentiation score or mass distributions compared to those of DF AB-62208 and DF AB- 54880 (KS test, p<Q.001). Individual cells are colored according to their senescence signature score. Blue shaded region is the putative zone of compatibility for RAS mutations and developmental state. FIG. 24G - Mass distributions for leukemia cells at MRD from DF AB- 25157 (sensitive to combination losmapimod) and DF AB-62208 (sensitive to combination fostamatinib). ** / ?<0.001 from paired Wilcoxon test.

[0091] FIG. 25A-25D - In vivo PDX Phase Il-like trial outcomes. FIG. 25A - Serial peripheral blood BCR::ABL1 qRT-PCR measurements from PDX model DF AB-13601 treated with Ponatinib daily (40mg / kg / day); each line represents an individual mouse. FIG. 25B - Key trial events and outcomes for each mouse on Phase Il-like trial, grouped by treatment arm. Complete response indicates <4% peripheral blood circulating blasts detected via flow cytometry; partial response indicates reduced peripheral blood blasts compared to pretreatment but >1% involvement; durable response indicates complete remission past 120 days on therapy. FIG. 25C - Competing risks model comparing progression and non-progression mortality in mice by treatment arm; p-values from a Cox regression analysis indicated for differences in progression and non-progression outcomes between treatment arms. FIG. 25D - Hazard ratios comparing pre-clinical risk factors for progression free survival in PDX mice (see Methods). Significant shifts (p<0.05 from Cox regression analysis) annotated in red. Median hazard ratios plotted with error bars representing ±1 quartile; "N"=number of mice; "HR"=hazard ratio; "CI"=confidence interval.

[0092] FIG. 26A-26C - Emergent patterns in BCR::ABL1 B-ALL mutation acquisition on TKI. FIG. 26A - Mutational alterations of individual PDX mice on TKI therapeutic regimen, grouped by disease stage and annotated by treatment arm ("Tx Arm"). Treatment emergent mutations indicated when mice from the same PDX line were profiled at pretreatment. Summary of grouped RAS or ABL pathway mutations included below. Mice are annotated for prior TKI exposure. "MRD"=minimal residual disease; "TFs"=transcription factors. Alteration details additionally reported in Table S3. FIG. 26B - Change in the fraction of mice on each Phase Il-like trial treatment arm that harbor mutations between progression and pretreatment. Genes along the ABL and RAS pathways are annotated in turquoise and magenta, respectively. FIG. 26C - Change in average VAF of PDX lines at progression for mutations along the ABL or RAS pathways compared to paired pretreatment tumors. Error bars indicate +1 standard deviation from the plotted mean AVAF.

[0093] FIG. 27A-27C - Intratumoral cNMF reveals developmentally convolved gene coexpression. FIG. 27A - cNMF program z-scored gene spectra for the top 30 metaprogram (mGEP) genes across all intratumoral gene expression programs (GEPs; Table S4); individual GEPs are annotated by PDX or Patient ID to show mGEP consensus across multiple donors. FIG. 27B - Representative heatmaps demonstrating intratumoral GEPs for one PDX tumor (DF AB-251574A0) and one patient tumor (BIAB-16768 Pretreatment). Known, healthy B cell lineage marker genes are annotated for each GEP. FIG. 27C - Pearson correlation of GEP module score and random forest (RF) classification probabilities. Bottom color track indicates the donor where each individual GEP was identified.

[0094] FIG. 28A-28F - Generation of healthy human bone marrow scRNA-seq dataset. FIG. 28A - Healthy human bone marrow samples (n = 7) were flow sorted into live bulk, CFU- L (colony-forming unit low; progenitor), Prim-B, Pro-B, Pre-Pre-B, Pre-B, and Immature B populations for scRNA-seq profiling (see Methods). FIG. 28B - Proportion of each cell type identified from the bulk (gray) or flow sorted-fraction (green). FIG. 28C - Force-directed graph (FDG) projection of healthy human bone marrow annotated by hematopoietic cell types (n=13,643 cells). FIG. 28D - Dot plot of hematopoietic cell type marker genes. Color denotes scaled average expression; size denotes percent expression in each scRNA-seq cell type population. FIG. 28E - FDG projection of healthy human bone marrow, annotated by donor. FIG. 28F - Donor fractional contribution to each cell type population.

[0095] FIG. 29A-29E - Random Forest Classifier accurately classifies healthy and Ph+ ALL single-cell transcriptomes. FIG. 29A - 10-fold cross-validation of each healthy reference cell type during RF training. FIG. 29B - Top 200 RF features ranked by permuted feature importance, grouped by healthy reference cell type (randomly down-sampled n=100 cells). FIG. 29C - Receiver Operating Characteristics (ROC) curves for RF classification of test scRNA-seq bone marrow dataset; area under the ROC curve (AUC) values listed in inset for each cell type. FIG. 29D - Shannon Diversity Index (SDI) of classification probabilities versus number of unique molecular identifier (UMI), number of genes, and percent mitochondrial transcripts for all leukemic cells. Cells removed from analysis due to highest non-B cell lineage classification are outlined in red and colored by misclassified cell type. Significant shifts in distribution between non-B lineage and B-lineage single-cells, as defined by a Kolmogorov- Smirnov (KS) test, reported (* / ?<0.001). FIG. 29E - Pearson correlation over gene expressionof top 2,000 highly-variable genes from healthy reference dataset across healthy and malignant hybrid cell type subpopulations.

[0096] FIG. 30A-30D - Defining Ph+ ALL developmental tumor hybrid populations. FIG. 30A - Developmental hybrid signatures defined by the top 30 genes correlated to RF prediction scores for each normal B-lineage cell type (Table S5). Average expression of signature genes across leukemic hybrid populations. FIG. 30B - Classification of leukemic hybrid populations based on random forest (RF) classification probabilities and hybrid signatures (see Methods). RF prediction probabilities, cycling or quiescent status, and PDX line or Patient ID annotated for each cell. FIG. 30C - Leukemic hybrid subpopulations projected onto RF prediction probability axes, as in Figure 20H. Densities of leukemia cells from each hybrid population projected over the landscape of all leukemia cells in the scRNA-seq dataset (plotted in grey). FIG. 30D - Scaled in silico predicted transcription factor (TF) activity over genes associated with developmental hybrid gene signatures (see Methods). Scaled TF activity scores shown in human reference samples (green) and PDX lines at pretreatment (grey) and progression (red), subset to TFs whose predicted activity scale with HSC, Pre-B, and Immature B RF classification probabilities in leukemic cells. Healthy reference Pre-BI and Pre-B II populations plotted independently within Pre-B.

[0097] FIG. 31A-31D - Transcriptional and immunophenotype shifts on therapy. FIG. 31A - Hybrid scRNA-seq population distributions for each profiled pretreatment and progression PDX mouse, annotated by treatment arm and time on treatment. FIG. 31B - Flow sorting gating strategy for B cell progenitor populations on a representative healthy human umbilical cord blood sample, PDX pre-treatment tumor, and PDX progression tumor (representative PDX=CB AB-75914). FIG. 31C - Fraction representation of PDX pretreatment and progression tumors across immunophenotyped B cell progenitor-like populations. Individual tumor immunophenotyped population fractions plotted as points; bars represent average tumor fraction within each immunophenotyped population at pretreatment or progression time points, including error bars for ±1 standard deviation. Surface markers used for flow gating of each population, as shown in (B), annotated below. FIG. 31D - Fraction of PDX tumor at progression of each immunophenotyped B cell progenitor-like population, grouped by mutation status at progression; bars represent average tumor fraction, with error bars for ±1 standard deviation. Significant p-values from Dirichlet regression noted; ** / ?<0.01 and * / ?<0.001.

[0098] FIG. 32A-32E - Random Forest (RF) Classifier recovers developmental structure in Smart-Seq2 single-cell transcriptomes. FIG. 32A - Proportion of RF cell type classifications across all Smart-Seq2 (SS2) healthy and leukemic cells. FIG. 32B - Single-cells ordered by RF prediction probabilities from progenitor cell types to differentiated B cell types, and annotated by flow sort gate (as in Figure 28A). Below, scaled expression of the top 10 RF prediction-correlated genes in developmentally-ordered healthy cells. FIG. 32C - Genes correlated to Pre-B RF prediction (x-axis) and genes correlated to Progenitor RF prediction are negatively correlated with each other; rho and p-value from Pearson correlation noted. Colored points represent the top 30 progenitor and Pre-B correlated genes used to define the SS2 developmental spectrum. FIG. 32D - Leukemic SS2 single-cells ranked by Progenitor-like score, annotated by B cell lineage RF prediction probabilities. Below, scaled expression of top 30 Progenitor-like and PreB-like signature genes (Table S7). FIG. 32E - Pearson crosscorrelation of RF cell type-correlated gene signature scores derived from Seq-Well and SS2 show cross-modality concordance. For clarity, SS2 signatures are hereafter referenced as "HSC-hyb" for Progenitor-like scores, and "PreB-hyb" for PreB-like scores.

[0099] FIG. 33A-33H - SS2 enables co-detection of mutations and transcriptome in leukemic single-cells. FIG. 33A - Summary of recurrently-identified RAS-pathway and ABL- pathway mutation loci from bulk targeted sequencing across PDX lines that were aligned for mutation detection in SS2 FASTQs (see Methods; Table S3). FIG. 33B - For genes with recurrently-identified mutations, Pearson correlation of average gene expression and normalized mutation-locus detection rate (either mutant or wild-type reads). FIG. 33C - Mutant and wild-type transcripts detected in SS2 single-cell transcriptomes from three representative PDX tumors; detected mutant transcript frequency in single-cells matched bulk VAF. FIG. 33D - Single-cell CNV profiles across each PDX line, including instances of CNV subclonal heterogeneity, paired with SS2-detected SNVs. FIG. 33E - SS2 single-cells within each profiled PDX line ordered by HSC-hyb expression scores, as defined in Figure 32D. Cycling status, CNV detection, and detected mutant and wild-type transcripts are annotated. Co-mutant indicates single-cells where RAS and ABL pathway mutations were detected. FIG. 33F - t-SNE projection of SS2 single-cells from representative PDX lines CBAB-12402, DF AB-62208, and DFAB-25157, colored by treatment time point. Number of SS2-profiled cells and mice at each time point denoted (n=cells, x=mice). FIG. 33G - All RAS-pathway mutant leukemic single-cells grouped by three treatment timepoints, annotated by KRAS orNRAS mutant transcript detection, and ordered by HSC-hyb signature scores within SS2 singlecells from DFAB-25157 and non-DFAB-25157 PDX lines demonstrates association between senescence-like and HSC-hyb gene expression scores across PDX lines and treatment stages. FIG. 33H - Cells from DFAB-25157 and DF AB-62208 at MRD and Progression, plot along fitness quadrants as defined in Figure 22H, with RAS-mutant leukemia cells annotated in red.

[0100] FIG. 34A-34D - Targeting integrative cell states enhances remission. FIG. 34A - All MRD single-cells ordered by Pre-BCR Signaling MRD state scores. CNV and SNV mutation status annotated for each cell, along with cycling and quiescent status. C- values reported from Fisher exact test comparing abundance of AV S'-mutant, quiescent, and / or cycling MRD cells with dominant Stress / Autophagy ("Stress / Auto.") expression scores to those with dominant Pre-BCR Signaling expression scores. FIG. 34B - Correlation between Stress / Autophagy and HSC-hyb gene expression, versus Pre-BCR signaling and PreB-hyb gene expression. Cycling cells annotated in red. FIG. 34C - Pathway enrichment false discovery rate (FDR) q-values for the top 100 genes in the Stress / Autophagy MRD state. FIG. 34D - Boxplot of relative MRD program (Pre-BCR Signaling - Stress / Autophagy) in MRD cells from DFAB-25157 and DF AB-62208; single-cell scores from each PDX-line plotted as individual points.

[0101] FIG. 35A-35C - Mass correlates with developmental states and cell cycle. FIG. 35A - Mass of healthy reference SS2 cells, binned by random forest-classified cell type and annotated by cell-type marker gene expression. Mean mass for each cell type plotted as a line. FIG. 35B - Force directed graph (FDG) visualization of healthy SS2 cells, annotated by cell type (top) and by cell mass (bottom); dot size indicates cell mass. FIG. 35C - Mass-correlated genes in healthy SS2 cells on the x-axis, versus the difference between genes correlated with RF progenitor and Pre-B cell types in healthy SS2 cells on the y-axis. Colored points denote marker genes for each cell type. R and p-value denote Pearson correlation between x- and y- axis indicated gene correlations.

[0102] The figures herein are for illustrative purposes only and are not necessarily drawn to scale.DETAILED DESCRIPTION OF THE EXAMPLE EMBODIMENTSGeneral Definitions

[0103] Unless defined otherwise, technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. Definitions of common terms and techniques in molecular biology may be found in Molecular Cloning: A Laboratory Manual, 2ndedition (1989) (Sambrook, Fritsch, and Maniatis); Molecular Cloning: A Laboratory Manual, 4thedition (2012) (Green and Sambrook); Current Protocols in Molecular Biology (1987) (F.M. Ausubel et al. eds.); the series Methods in Enzymology (Academic Press, Inc.): PCR 2: A Practical Approach (1995) (M.J. MacPherson, B.D. Hames, and G.R. Taylor eds.): Antibodies, A Laboratory Manual (1988) (Harlow and Lane, eds.): Antibodies A Laboratory Manual, 2ndedition 2013 (E.A. Greenfield ed.); Animal Cell Culture (1987) (R.I. Freshney, ed.); Benjamin Lewin, Genes IX, published by Jones and Bartlet, 2008 (ISBN 0763752223); Kendrew et al. (eds.), The Encyclopedia of Molecular Biology, published by Blackwell Science Ltd., 1994 (ISBN 0632021829); Robert A. Meyers (ed.), Molecular Biology and Biotechnology: a Comprehensive Desk Reference, published by VCH Publishers, Inc., 1995 (ISBN 9780471185710); Singleton etal., Dictionary of Microbiology and Molecular Biology 2nd ed., J. Wiley & Sons (New York, N.Y. 1994), March, Advanced Organic Chemistry Reactions, Mechanisms and Structure 4th ed., John Wiley & Sons (New York, N.Y. 1992); and Marten H. Hofker and Jan van Deursen, Transgenic Mouse Methods and Protocols, 2ndedition (2011).

[0104] Titles or subtitles may be used in the specification for the sole convenience of the reader but are not intended to influence the scope of the present disclosure or to limit any aspect of the disclosure to any subsection, subtitle, or paragraph.

[0105] As used herein, the singular forms “a”, “an”, and “the” include both singular and plural referents unless the context dictates otherwise.

[0106] The term “optional” or “optionally” means that the element or step that follows the term may or may not occur, and the description is intended to cover embodiments that include the element or step or do not include it.

[0107] The recitation of numerical ranges by endpoints includes all numbers and fractions subsumed within the respective ranges and the recited endpoints.

[0108] The terms “about” or “approximately” as used herein when referring to a measurable value, such as a parameter, an amount, a temporal duration, and the like, are meantto encompass variations of and from the specified value, such as variations of + / - 10% or less, + / -5% or less, + / -1% or less, and + / -0.1% or less of and from the specified value, insofar such variations are appropriate to perform in the disclosure. It is understood that the value to which the modifier “about” or “approximately” refers is also specifically and preferably disclosed.

[0109] As used herein, “tissue” refers to an aggregation of morphologically similar cells and associated intercellular matter, i.e., extracellular matrix, acting together to perform one or more specific bodily functions. In an embodiment, tissues fall into one of four basic types: muscle, nerve, epidermal, and connective. In an embodiment, a tissue is substantially solid, e.g., cells within the tissue are strongly associated with one another to form a multicellular solid tissue. In an embodiment, a tissue can be a tumor. In an embodiment, the tissue can be a lymph node. In an embodiment, tissue is substantially non-solid, e.g., cells within the tissue are loosely associated or not physically associated with one another but may be found in the same space, bodily fluid, etc. For example, blood cells are considered a tissue in non-solid form.

[0110] The terms “subject,” “individual,” and “patient” are used interchangeably herein to refer to a vertebrate, preferably a mammal, more preferably a human. Non-limiting examples of mammals include members of the human, simian, equine, porcine, bovine, rattus, murine, canine, and feline species. Mammals also include, but are not limited to, farm animals, sport animals, and pets. Tissues, cells, and the progeny of a biological entity obtained in vivo or cultured in vitro are also encompassed.[OHl] As used herein, “tumor” refers to all neoplastic cell growth and proliferation, whether malignant or benign and all pre-cancerous and cancerous cells and tissues. “Neoplastic,” as used herein, refers to any form of dysregulated or unregulated cell growth, whether malignant or benign, resulting in abnormal tissue growth. Thus, “neoplastic cells” include malignant and benign cells having dysregulated or unregulated cell growth.

[0112] As used herein, the terms “treat,” “treatment,” and “treating” refer to treatments wherein the object is to reverse, alleviate, ameliorate, inhibit, slow down, or stop the progression or severity of a cancer. The term “treating” includes reducing or alleviating at least one adverse effect or symptom of cancer. Treatment is generally “effective” if one or more symptoms or clinical markers are reduced. Alternatively, treatment is “effective” if cancer progression is diminished or halted. That is, “treatment” includes not just the improvement of symptoms or markers but also a cessation of, or at least slowing of, progress or worsening ofsymptoms compared to what would be expected in the absence of treatment. Beneficial or desired clinical results include but are not limited to alleviation of one or more symptom(s), dimini shm ent of the extent of the cancer, stabilized (i.e., not worsening) state of the cancer, delay or slowing of cancer progression, amelioration or palliation of the cancer, remission (whether partial or total), and / or decreased mortality, whether detectable or undetectable. The term “treatment” of a disease also includes relief from the symptoms or side effects of the cancer (including palliative treatment). The term “sensitivity” or “sensitive” when made about treatment with a compound is a relative term that refers to the degree of effectiveness of the compound in lessening or decreasing the progress of a tumor or the disease being treated. For example, the term “increased sensitivity” when used about treating a cell or tumor in connection with a compound refers to an increase of at least about 5%, or more, in the effectiveness of the tumor treatment.

[0113] As used herein, and unless otherwise specified, the term “therapeutically effective amount” of a compound is sufficient to provide a therapeutic benefit in treating or managing cancer or to delay or minimize one or more symptoms associated with cancer. A therapeutically effective amount of a compound means an amount of therapeutic agent, alone or in combination with other therapies, which provides a therapeutic benefit in the treatment or management of the cancer. The term “therapeutically effective amount” can encompass an amount that improves overall therapy, reduces or avoids symptoms or causes of cancer, or enhances the therapeutic efficacy of another therapeutic agent. The term also refers to the amount of a compound that is sufficient to elicit the biological or medical response of a biological molecule (e.g., a protein, enzyme, RNA, or DNA), cell, tissue, system, animal, or human, which a researcher, veterinarian, medical doctor, or clinician are seeking.

[0114] The term “responsiveness” or “responsive” when used in reference to a treatment refers to the effectiveness of the treatment in lessening or decreasing the symptoms of a disease, e.g., ALL, being treated. For example, the term “increased responsiveness” when used in reference to a treatment of a cell or a subject refers to an increase in the effectiveness in lessening or decreasing the symptoms of the disease compared to a reference treatment (e.g., of the same cell or subject, or a different cell or subject) when measured using any methods known in the art. In an embodiment, the increase in effectiveness is at least about 5%, at least about 10%, at least about 20%, at least about 30%, at least about 40%, or at least about 50%.

[0115] As used herein, the terms “effective subject response,” “effective patient response,” and “effective patient tumor response” refer to any increase of a therapeutic benefit to the patient. An “effective patient tumor response” can be, for example, about 5%, about 10%, about 25%, about 50%, or about a 100% decrease in the tumor's progress rate. An “effective patient tumor response” can be about 5%, about 10%, about 25%, about 50%, or about a 100% decrease in the physical symptoms of a cancer. An “effective patient tumor response” can also be, for example, about 5%, about 10%, about 25%, about 50%, about 100%, about 200%, or more increase in the response of the patient, as measured by any suitable means, such as gene expression, cell counts, assay results, tumor size, etc.

[0116] When a range of values is listed herein, it is intended to encompass each value and subrange within that range. For example, “1-5 ng” or a range of from “1 ng to 5 ng” is intended to encompass 1 ng, 2 ng, 3 ng, 4 ng, 5 ng, 1-2 ng, 1-3 ng, 1-4 ng, 1-5 ng, 2-3 ng, 2-4 ng, 2-5 ng, 3-4 ng, 3-5 ng, and 4-5 ng.

[0117] An improvement in the cancer or cancer-related disease can be characterized as a complete or partial response. “Complete response” refers to an absence of clinically detectable disease with normalization of any previously abnormal radiographic studies, bone marrow, cerebrospinal fluid (CSF), or abnormal monoclonal protein measurements. “Partial response” refers to at least about 10%, about 20%, about 30%, about 40%, about 50%, about 60%, about 70%, about 80%, or about 90% decrease in all measurable tumor burden (i.e., the number of malignant cells present in the subject, or the measured bulk of tumor masses or the quantity of abnormal monoclonal protein) in the absence of new lesions. The term “treatment” contemplates both a complete and a partial response.

[0118] The term “likelihood” generally refers to an increase in the probability of an event. The term “likelihood,” when used in reference to the effectiveness of a subject’s tumor response, generally contemplates an increased probability that the rate of tumor progress or tumor cell growth will decrease. The term “likelihood,” when used in reference to the effectiveness of a patient tumor response, can also generally mean the increase of indicators, such as mRNA or protein expression, that may increase the progress in treating the tumor.

[0119] The term “predict” generally means to determine or tell in advance. When used to “predict” the effectiveness of a cancer treatment, for example, the term “predict” can mean that the likelihood of the outcome of the cancer treatment can be determined at the outset, before the treatment has begun, or before the treatment period has progressed substantially.

[0120] The term “monitor,” as used herein, generally refers to the overseeing, supervising, regulating, watching, tracking, or surveillance of an activity. For example, “monitoring the effectiveness of a compound” refers to tracking the effectiveness of treating cancer in a patient or a tumor cell culture. Similarly, the term “monitoring,” when used in connection with patient compliance, either individually or in a clinical trial, refers to the tracking or confirming that the patient is taking a drug being tested as prescribed. For example, the monitoring can be performed by following the expression of mRNA or protein biomarkers. In addition, “monitoring” can be used herein in the context of gene expression changes or for new mutations in tissues such as tumor tissues during various stages of disease, including remission, minimal residual disease, or relapse.

[0121] The term “regulate,” as used herein, refers to controlling the activity of a molecule or biological function, such as enhancing or diminishing the activity or function.

[0122] The term “refractory” or “resistant” refers to a circumstance where patients, even after intensive treatment, have residual cancer cells (e.g., leukemia or lymphoma cells) in their lymphatic system, blood, and / or blood-forming tissues (e.g., marrow).

[0123] As used herein, “remission” refers to a state with a significant reduction or complete absence of leukemia symptoms and signs. This state is characterized by normalizing blood cell counts and the lack of detectable leukemia cells in the bone marrow and blood, as determined by diagnostic tests. Remission indicates that the treatment has effectively controlled the disease but does not necessarily mean the leukemia has been cured. First-line therapy and subsequent treatments aim to achieve and maintain remission, improving the patient's quality of life and survival prospects.

[0124] The term “relapse” refers to a situation in which patients who have had a remission of cancer after therapy have a return of cancer cells (e.g., leukemia or lymphoma cells) in their lymphatic system, blood, and blood-forming tissues (e.g., bone marrow) and a decrease in normal blood cells. Relapse can be detected through various diagnostic methods, including identifying specific cell states or transcriptional programs indicative of leukemia.

[0125] A “biological marker” or “biomarker” is a molecule whose detection indicates a particular biological state, such as, for example, the presence of cancer. In further embodiments, biomarkers can be determined individually. In other embodiments, several biomarkers can be measured simultaneously. In an embodiment, detecting a biomarker or a change in the amount of a biomarker can indicate a change in the level of gene expression thatcorrelates with the risk or progression of a disease or with the susceptibility of the disease to a given treatment. The biomarker can be a nucleic acid, such as mRNA or cDNA, in further embodiments.

[0126] In an embodiment, detecting a biomarker or a change in the amount of a biomarker can indicate a change in the polypeptide or protein expression level that correlates with the risk or progression of a disease or a patient's susceptibility to treatment. In further embodiments, the biomarker can be a polypeptide, protein, or fragment thereof. The relative level of specific proteins can be determined by methods known in the art. For example, antibody -based methods, such as an immunoblot, enzyme-linked immunosorbent assay (ELISA), or other methods can be used.

[0127] The term “sample,” as used herein, relates to a material or mixture of materials, typically, although not necessarily, in fluid form, containing one or more components of interest.

[0128] As used herein, a “biological sample” may contain whole cells and / or live cells and / or cell debris. The biological sample may include (or be derived from) a “bodily fluid” or tissue sample. The present disclosure encompasses embodiments wherein the bodily fluid is selected from amniotic fluid, aqueous humour, vitreous humour, bile, blood serum, breast milk, cerebrospinal fluid, cerumen (earwax), chyle, chyme, endolymph, perilymph, exudates, feces, female ejaculate, gastric acid, gastric juice, lymph, mucus (including nasal drainage and phlegm), pericardial fluid, peritoneal fluid, pleural fluid, pus, rheum, saliva, sebum (skin oil), semen, sputum, synovial fluid, sweat, tears, urine, vaginal secretion, vomit and mixtures of one or more thereof. Biological samples include cell cultures, bodily fluids, and cell cultures from bodily fluids. Bodily fluids may be obtained from a mammal organism, for example, by puncture or other collecting or sampling procedures. In an embodiment, the sample can be cell- free (e.g., cell-free DNA or cfDNA).

[0129] Various embodiments are described hereinafter. It should be noted that the specific embodiments are not intended as an exhaustive description or as a limitation to the broader aspects discussed herein. One aspect described in conjunction with a particular embodiment is not necessarily limited to that embodiment and can be practiced with any other embodiment s). Reference throughout this specification to “one embodiment,” “an embodiment,” and “an example embodiment” means that a particular feature, structure, or characteristic described inconnection with the embodiment is included in at least one embodiment of the present disclosure.OVERVIEW

[0130] Mutations in residual leukemia cells are believed to contribute to therapeutic resistance in oncogene-addicted cancers. However, comprehensive assessments of how variations across different molecular layers co-evolve and impact treatment are limited. As demonstrated herein, the same mutation in cancer cells present at remission can lead to divergent outcomes if it co-occurs with different transcriptional states, establishing the importance of cell state in determining the fitness of the residual cell and its ability to contribute to relapse. Methods disclosed herein directly target cell states present during remission to enhance response compared to interventions informed by mutations alone and provide a rapid biophysical readout of cell states with corresponding genotypes.METHODS OF TREATING LEUKEMIA

[0131] The disclosed methods center on a personalized approach to treating leukemia by determining the specific cellular and transcriptional characteristics of leukemic cells in a subject. In an embodiment, the method incorporates a dual-phase diagnostic and treatment strategy, including detection and treatment phases. In one aspect, embodiments disclosed herein are directed to methods of treating leukemia that comprise detecting in a subject suffering from leukemia a Pre-B cell receptor (Pre-BCR) or a stress-autophagy state and then administering to the subject a BCR signaling pathway inhibitor or a tyrosine kinase inhibitor (TKI) in combination with a SYK inhibitor if a pre-BCR signaling state is detected, or administering to the subject a TKI inhibitor in combination with a p38 inhibitor if a stress autophagy program is detected.

[0132] The term “leukemia” refers to malignant neoplasms of the blood-forming tissues. The leukemia may include, but is not limited to, chronic lymphocytic leukemia, chronic myelocytic leukemia, acute lymphoblastic leukemia, acute myeloid leukemia, and acute myeloblastic leukemia. The leukemia can be relapsed, refractory, or resistant to conventional therapy. In an embodiment, the leukemia is oncogene-addicted. An oncogene-addicted leukemia refers to a subtype of leukemia that exhibits a pathological reliance on one or a few oncogenes for its initiation, survival, and proliferation. The term “oncogene addiction” refers to the concept that the cancerous behavior of leukemic cells is predominantly driven by specific oncogenes, which may encode for aberrant or overactive proteins that signal cells to grow anddivide uncontrollably. In an embodiment, the leukemia is an acute lymphoblastic leukemia. In an embodiment, the oncogene-addicted leukemia is BCR-ABL1 B-cell acute lymphoblastic leukemia (B-ALL). In an embodiment, the subject suffering from leukemia may be a subject who is in relapse following treatment with induction therapy.

[0133] In an embodiment, the subject relapses following an initial induction therapy. “Induction therapy” in the context of leukemia is the first phase of treatment given after diagnosis. Its primary goal is to induce remission, decreasing the number of leukemia cells to undetectable levels and allowing the bone marrow to function normally again. This phase involves intensive chemotherapy and, in some types of leukemia, like BCR-ABL1 -positive acute lymphoblastic leukemia (ALL), may include targeted therapies such as tyrosine kinase inhibitors (TKIs). The regimen for induction therapy is often aggressive and may require hospitalization to manage both the therapy and its potential complications, like infections due to reduced immunity. Therapies for leukemia are outlined in the Clinical Practice Guidelines in Oncology: National Comprehensive Cancer Network (https: / / www.nccn.org / guidelines), which are incorporated herein by reference.Detecting Cell States

[0134] The detecting step may comprise detecting a cell state in residual cells present in a subject after an initial treatment or after an initial successful treatment into remission. The cell type may be one or more immune cells. The one or more immune cells may be obtained from any suitable biological sample from the subject, including a bone marrow biopsy. A “cell state” refers to the condition or phase that a cell is in at a given time based on its cell cycle, differentiation status, and physiological condition. Various factors, including genetic expression, metabolic activity, and external environmental conditions, may determine this state. In an embodiment, the cell state to be detected is a pre-BCR or a stress / autophagy cell state. Cell states can be detected through the detection of transcriptional programs or the measurement of biophysical properties. Biophysical properties can correlate with underlying transcriptional states and be used as an alternative measurement to detect cell states. Singlecell transcriptomics can be expensive and requires advanced technical skills for measurement and interpretation. Biophysical measurements provide a metric that can integrate complex transcriptional information from low-input samples to rapidly determine the leukemic developmental state.Detecting Cell States Via Transcriptional Programs

[0135] In an embodiment, the cell state may be detected via the detection of a transcriptional program. Changes in gene expression define transcriptional programs and can be detected using suitable methods for identifying changes in transcription. These methods may include, but are not limited to, hybridization, amplification, and / or sequencing methods. In an embodiment, transcriptional programs are detected using next-generation sequencing techniques, such as RNA sequencing (RNA-seq). The cell state may be detected by characterizing one or more biophysical features in an embodiment. Examples of biophysical measurements that may be used include size, mass, morphology, mechanical properties, or cellular stiffness.Detecting Cell States Via Measuring Single-Cell Biophysical Properties

[0136] Single-cell biophysical properties, including electrical properties (cytoplasm conductivity and specific membrane capacitance), geometric properties (cell size), and cell mass, are closely related to gene and protein expression and are widely used in cell type classification and state evaluation. Compared with commonly used cellular genetic and protein detection methods, such as polymerase chain reaction and mass spectrometry, electrical measurement methods are rapid, simple, and less invasive. Commonly used electrical measurement methods include patch clamping, electrical impedance spectroscopy (EIS), electrorotation (ROT), and electrophoresis (DEP).

[0137] The patch-clamp approach characterizes the activities of cellular ion channels by sucking a portion of the cell membrane into a micropipette tip to form a high electrical resistance seal, enabling quantification of the specific membrane capacitance of a single cell. However, the patch clamp method is time-consuming, inconvenient, and somewhat invasive to cells.

[0138] Electrical impedance spectroscopy (EIS) measures the electrical impedance of a cell by using electrodes, where a frequency-dependent signal is applied and the electrical responses are recorded. It can be divided into ECIS (Electric Cell-substrate Impedance Sensing) and MIC (Microfluidic Impedance Cytometry). ECIS measures the electrical impedances of adherent cells by configuring microelectrodes on the substrate. MIC is used for measuring suspended cells. When single cells pass through a microchannel, the electrodes alongside the fluidic channel detect the impedance variations. EIS methods can only extract the equivalent capacitance and resistance of the cell, rather than its electrical properties, due tothe problems of current leakage and the need for appropriate models. A modified MIC was recently reported to measure cellular electrical parameters of specific membrane capacitance and cytoplasm conductivity when the cell was squeezed into a constricted channel to fix the cell shape and reduce leakage current. However, the method probably caused a blockage of the microchannel, and the cell was squeezed rather than in its natural state. Additionally, an on- chip capacitive detection method was proposed to measure the capacitance of a single Escherichia coli (E. coli) cell from drinking water.

[0139] Electrorotation (ROT) measures the rotation speed of single cells under a rotational electric field. The electrical properties of cells, including tumor cells, blood cells, and stem cells, can be extracted by fitting theoretical rotation spectra to experimental data. Rotation spectra refer to the cellular rotation rates as a function of the frequencies of the applied fields. However, cells tend to deviate from their initial positions during rotation, which changes the rotation speed.

[0140] Dielectrophoresis (DEP) is a motion imparted on electrically neutral but polarized particles subjected to non-uniform electric fields. Particles moving toward low electric field region are called “negative DEP (nDEP).” Particles moving toward a high electric field region are called “positive DEP (pDEP).” The DEP technology has been widely used in trapping, manipulating, and separating cells due to its easy operation, less invasive and label-free. There are two main methods to measure cell electrical properties using DEP. One is to measure DEP crossover frequency (CF), i.e., a frequency occurring on the transition between pDEP and nDEP, which is related to the cell’s electrical properties. The DEP-CF method suffers from limitations easily interfered with by external forces, is time-consuming, and has special requirements for solution conductivity.Cell mass

[0141] Cell mass can also be determined using suspended microchannel resonators. Cells are weighed in real-time with the suspended microchannel resonator (SMR) as they flow through a hollow cantilever. In addition to weighing particles with femtogram precision, the SMR can measure mass density with a resolution of 10-4 g / mL. This is possible since the particle's mass difference concerning the displaced fluid determines the microchannel resonant frequency. Thus, the particle's density is determined by measuring its mass in two fluids of different densities.

[0142] The SMR can measure the growth of single cells and the absorption of biomolecules to microchannel walls. Scaling the current SMR design tenfold will result in a thousand-fold mass resolution (attogram) improvement, enabling single viruses to be weighed with high precision. See, e.g., Ko J et al., Cellular and biomolecular detection based on suspended microchannel resonators. Biomed Eng. Lett. 2021 Sep 12; 11(4):367-382, incorporated herein by this reference.

[0143] Microfluidic Mass Sensors: These devices can measure the mass of single cells with high precision. Microfluidic channels guide cells over resonating sensors that change frequency in response to the mass of the passing cell. This non-invasive method allows for the real-time monitoring of cell mass changes.

[0144] Quartz Crystal Microbalance (QCM): This technique uses the principle of quartz crystal oscillation frequency change upon mass addition. Cells adhering to the quartz surface cause a measurable frequency shift proportional to their mass. QCM is sensitive and can measure mass changes in real time, but it requires cells to adhere to the crystal surface, limiting its application to specific cell types.

[0145] Optical Methods: Recent advancements include optical interferometry and holography approaches that can measure the mass and density of cells in suspension. These methods rely on the phase shift of light passing through a cell, which is correlated with the cell's mass and refractive index. Optical methods are non-invasive and suitable for live cells, offering the advantage of measuring cells without needing attachment or labeling.

[0146] Centrifugal Methods: Techniques like buoyant mass measurement use centrifugation to determine cell mass based on buoyancy. Cells are suspended in a fluid of known density and centrifuged to reach equilibrium at a position where their buoyant density matches that of the surrounding medium. The position is then used to calculate cell mass.

[0147] Atomic Force Microscopy (AFM): Although primarily used for imaging, AFM can also measure the mass of cells and other microscopic particles. Researchers can estimate the cell's mass by measuring the force required to deflect the AFM tip when it contacts a cell. This technique offers high spatial resolution but is more complex and time-consuming than other methods.

[0148] Applicants demonstrate herein that the progenitor-cell state can be identified as cells with a higher cell mass, and the pre-B cell-like state can be identified as cells with a lower cell mass. In an embodiment, a higher cell mass is between 15 to 25pg, between 16 to 24pg,between 17 to 23pg, between 18 to 22pg, or between 19 to 21pg. In an embodiment, a lower cell mass is between 10 to 15pg, between 11 to 14pg, or between 12 to 13pg.Pre-BCR Cell State

[0149] The pre-BCR cell state in leukemic cells is characterized by a differentiation stage that resembles normal pre-B cells undergoing maturation in the bone marrow. This program involves a transcriptional profile indicative of active pre-BCR signaling, which is essential for pre-B cells' survival, proliferation, and further differentiation. Targeting the pre-BCR signaling pathway can be particularly effective in treating leukemia subtypes that rely on these signals for maintenance and progression.

[0150] In a pre-BCR cell state, leukemic cells express a functional Pre-BCR complex consisting of the immunoglobulin mu heavy chain paired with surrogate light chains and signaling components such as Ig-a / Ig-P heterodimers. This receptor engagement leads to a cascade of intracellular signaling events involving kinases like SYK and proteins in the B-cell linker (BLNK) pathway. The downstream effect of Pre-BCR signaling can influence genes regulating cell cycle progression, apoptosis, and metabolic processes, ultimately promoting cellular expansion and survival. Inhibition of this pathway could impair the survival mechanisms of leukemic cells, reduce their proliferative capacity, and potentially sensitize them to other therapeutic agents. In an embodiment, the Pre-BCR state is an aspect of the biology of pre-B cell leukemias and serves as a potential therapeutic target. By interfering with the molecular mechanisms underlying Pre-BCR signaling, new strategies can be developed to treat leukemias that exhibit this signaling program, potentially leading to better clinical outcomes.

[0151] In an embodiment, the pre-BCR cell state may be determined by a single cell measurement of the cell mass of a cell derived from a sample from the subject suffering from leukemia. In an embodiment, a cell mass between 15 to 25 pg, 16 to 25pg, 17 to 25 pg, 18 to 25 pg, 19 to 25 pg, 20 to 25 pg, 21 to 25 pg, 22 to 25 pg, 23-25 pg indicates a pre-BCR cell state. In an embodiment, a single cell mass of 15pg, 16pg, 17pg, 18pg, 19pg, 20pg, 21pg, 22pg, 23pg, 24pg, or 25pg indicates a pre-BCR cell state.

[0152] In an embodiment, the pre-BCR cell state is detected by detecting a pre-BCR transcriptional program. In an embodiment, detecting the pre-BCR cell transcriptional program comprises detecting one or more genes selected from IGLL1, VPREB3, TCL1A, UHRF1, H1F0, SOX11, RPS4Y1, VDAC1, CCND3, ARPP21, HMGN2, RPS26, CMTM8, IGJ,A0X2P, TMSB4X, FAM129C, SNURF, PTMA, LAT2, HIST1H3G, TP53INP1, SNRPN, NELLI, DNMT1, HMGB1, CBX1, PHGDH, LGALS9, EBF1, FLU, RPS4Y2, C16orf54, STMN1, ERGIC1, IRF2BP2, PAXS, E2F2, CD38, GNG7, P0U2AF1, UBE2T, MME, NUSAP1, D0K3, PVRIG, PIP4K2A, UCP2, and DEK.

[0153] In an embodiment, the pre-BCR cell state is detected by detecting a pre-BCR transcriptional program. In an embodiment, detecting the pre-BCR cell transcriptional program comprises detecting two or more genes selected from IGLL1, VPREB3, TCL1A, UHRF1, H1F0, SOX11, RPS4Y1, VDAC1, CCND3, ARPP21, HMGN2, RPS26, CMTM8, IGJ, AOX2P, TMSB4X, FAM129C, SNURF, PTMA, LAT2, HIST1H3G, TP53INP1, SNRPN, NELLI, DNMT1, HMGB1, CBX1, PHGDH, LGALS9, EBF1, FLU, RPS4Y2, C16orf54, STMN1, ERGIC1, IRF2BP2, PAXS, E2F2, CD38, GNG7, POU2AF1, UBE2T, MME, NUSAP1, DOK3, PVRIG, PIP4K2A, UCP2, and DEK.

[0154] In an embodiment, the pre-BCR cell state is detected by detecting a pre-BCR transcriptional program. In an embodiment, detecting the pre-BCR cell transcriptional program comprises detecting five or more genes selected from IGLL1, VPREB3, TCL1A, UHRF1, H1F0, SOX11, RPS4Y1, VDAC1, CCND3, ARPP21, HMGN2, RPS26, CMTM8, IGJ, AOX2P, TMSB4X, FAM129C, SNURF, PTMA, LAT2, HIST1H3G, TP53INP1, SNRPN, NELLI, DNMT1, HMGB1, CBX1, PHGDH, LGALS9, EBF1, FLU, RPS4Y2, C16orf54, STMN1, ERGIC1, IRF2BP2, PAXS, E2F2, CD38, GNG7, POU2AF1, UBE2T, MME, NUSAP1, DOK3, PVRIG, PIP4K2A, UCP2, and DEK.

[0155] In an embodiment, the pre-BCR cell state is detected by detecting a pre-BCR transcriptional program. In an embodiment, detecting the pre-BCR cell transcriptional program comprises detecting ten or more genes selected from IGLL1, VPREB3, TCL1A, UHRF1, H1F0, SOX11, RPS4Y1, VDAC1, CCND3, ARPP21, HMGN2, RPS26, CMTM8, IGJ, AOX2P, TMSB4X, FAM129C, SNURF, PTMA, LAT2, HIST1H3G, TP53INP1, SNRPN, NELLI, DNMT1, HMGB1, CBX1, PHGDH, LGALS9, EBF1, FLU, RPS4Y2, C16orf54, STMN1, ERGIC1, IRF2BP2, PAXS, E2F2, CD38, GNG7, POU2AF1, UBE2T, MME, NUSAP1, DOK3, PVRIG, PIP4K2A, UCP2, and DEK.

[0156] In an embodiment, the pre-BCR cell state is detected by detecting a pre-BCR transcriptional program comprising IGLL1, VPREB3, TCL1A, UHRF1, H1F0, SOX11, RPS4Y1, VDAC1, CCND3, ARPP21, HMGN2, RPS26, CMTM8, IGJ, AOX2P, TMSB4X, FAM129C, SNURF, PTMA, LAT2, HIST1H3G, TP53INP1, SNRPN, NELLI, DNMT1,HMGB1, CBX1, PHGDH, LGALS9, EBF1, FLU, RPS4Y2, C16orf54, STMN1, ERGIC1, IRF2BP2, PAXS, E2F2, CD38, GNG7, P0U2AF1, UBE2T, MME, NUSAP1, D0K3, PVRIG, PIP4K2A, UCP2, and DEK.Stress / Autophagy Cell State

[0157] The stress / autophagy cell state in leukemia refers to the cellular response that allows cancer cells to survive under metabolic stress and adverse environmental conditions. This program is characterized by activating a network of genes that facilitate autophagy. In this process, cells recycle their components to sustain metabolic needs and adapt to stressors such as nutrient deprivation, hypoxia, and treatment-induced damage. Therapeutically, targeting leukemia's stress / autophagy program can be beneficial, especially when combined with other treatments. Inhibitors that disrupt autophagy, such as chloroquine and its derivatives, or specific targeting of regulatory components of the autophagy pathway, have shown promise in preclinical studies and are being explored in clinical trials to enhance the efficacy of anticancer therapies and to overcome resistance mechanisms. The stress / autophagy cell state comprises transcriptionally enriched cells for senescence.

[0158] In an embodiment, the stress / autophagy cell state may be determined by a single cell measurement of the cell mass of a cell derived from a sample from the subject suffering from leukemia. In an embodiment, a cell mass is between 10 to 15 pg, 10 to 14pg, 10 to 13 pg, and 10 to 12pg. In an embodiment, a single cell mass of lOpg, 1 Ipg, 12pg, 13pg, orl4pg,

[0159] In an embodiment, the stress / autophagy cell state is detected by detecting a stress / autophagy transcriptional program. In an embodiment, detecting the stress / autophagy transcriptional program comprises one or more genes selected from DNAJB1, HSPA1A, HSP90AB1, HSPE1, HSPH1, DNAJA1, HSPA1B, NR4A1, HSPA8, CDKN1A, HBEGF, HSP90AA1, HSPB1, HSPA6, HSPD1, ZFAND2A, TUBB4B, EIF5, UBC, EIF1, DNAJB6, DDX3X, CHORDCI, JUN, ZC3HAV1, IER3, HERPUD1, IFRD1, BAG3, SAP18, HSA1, PMAIP1, BRD2, HSPA7, MAP1LC3B, SLC2A3, ZFAS1, TCP1, EIF4G2, ATF3, INSIGI, SOD2, SAT1, DEDD2, C6orf62, CACYBP, HEXIM1, EGR3, and CSRNP1.

[0160] In an embodiment, detecting the stress / autophagy transcriptional program comprises two or more genes selected from DNAJB1, HSPA1A, HSP90AB1, HSPE1, HSPH1, DNAJA1, HSPA1B, NR4A1, HSPA8, CDKN1A, HBEGF, HSP90AA1, HSPB1, HSPA6, HSPD1, ZFAND2A, TUBB4B, EIF5, UBC, EIF1, DNAJB6, DDX3X, CHORDCI, JUN, ZC3HAV1, IER3, HERPUD1, IFRD1, BAG3, SAP18, HSA1, PMAIP1, BRD2, HSPA7,MAP1LC3B, SLC2A3, ZFAS1, TCP1, EIF4G2, ATF3, INSIGI, SOD2, SAT1, DEDD2, C6orf62, CACYBP, HEXIM1, EGR3, and CSRNPE

[0161] In an embodiment, detecting the stress / autophagy transcriptional program comprises five or more genes selected from DNAJB1, HSPA1A, HSP90AB1, HSPE1, HSPH1, DNAJA1, HSPA1B, NR4A1, HSPA8, CDKN1A, HBEGF, HSP90AA1, HSPB1, HSPA6, HSPD1, ZFAND2A, TUBB4B, EIF5, UBC, EIF1, DNAJB6, DDX3X, CHORDCI, JUN, ZC3HAV1, IER3, HERPUD1, IFRD1, BAG3, SAP18, HSA1, PMAIP1, BRD2, HSPA7, MAP1LC3B, SLC2A3, ZFAS1, TCP1, EIF4G2, ATF3, INSIGI, SOD2, SAT1, DEDD2, C6orf62, CACYBP, HEXIM1, EGR3, and CSRNPE

[0162] In an embodiment, detecting the stress / autophagy transcriptional program comprises ten or more genes selected from DNAJB1, HSPA1A, HSP90AB1, HSPE1, HSPH1, DNAJA1, HSPA1B, NR4A1, HSPA8, CDKN1A, HBEGF, HSP90AA1, HSPB1, HSPA6, HSPD1, ZFAND2A, TUBB4B, EIF5, UBC, EIF1, DNAJB6, DDX3X, CHORDCI, JUN, ZC3HAV1, IER3, HERPUD1, IFRD1, BAG3, SAP18, HSA1, PMAIP1, BRD2, HSPA7, MAP1LC3B, SLC2A3, ZFAS1, TCP1, EIF4G2, ATF3, INSIGI, SOD2, SAT1, DEDD2, C6orf62, CACYBP, HEXIM1, EGR3, and CSRNPE

[0163] In an embodiment, detecting the stress / autophagy transcriptional program comprises DNAJB1, HSPA1A, HSP90AB1, HSPE1, HSPH1, DNAJA1, HSPA1B, NR4A1, HSPA8, CDKN1A, HBEGF, HSP90AA1, HSPB1, HSPA6, HSPD1, ZFAND2A, TUBB4B, EIF5, UBC, EIF1, DNAJB6, DDX3X, CHORDCI, JUN, ZC3HAV1, IER3, HERPUD1, IFRD1, BAG3, SAP18, HSA1, PMAIP1, BRD2, HSPA7, MAP1LC3B, SLC2A3, ZFAS1, TCP1, EIF4G2, ATF3, INSIGI, SOD2, SAT1, DEDD2, C6orf62, CACYBP, HEXIM1, EGR3, and CSRNPEDevelopmental Maturity and Treatment Resistance

[0164] In an embodiment, leukemic cells are characterized by their degree of developmental maturity along the normal B-cell lineage, wherein developmental maturity exists on a continuous spectrum that correlates with therapeutic resistance. In another embodiment, Philadelphia chromosome positive (Ph+) leukemic cells with mature B-cell developmental characteristics demonstrate reduced sensitivity to ABL inhibitors compared to cells with progenitor-like characteristics, and such mature B-cell developmental characteristics are determined using random forest classification against healthy bone marrow reference data to assign developmental stage probabilities.

[0165] In an embodiments, ABL inhibitor sensitivity is predicted prior to treatment initiation by evaluating the developmental maturity profile of leukemic cells from a subject, wherein leukemic cells classified as having mature B-cell signatures require enhanced combination therapy approaches beyond standard tyrosine kinase inhibitor monotherapy. In an embodiment, the developmental maturity assessment identifies patients likely to experience resistance to standard ABL inhibitor therapy, enabling consideration of alternative or combination treatment strategies, and cell lines or primary patient samples are evaluated for developmental staging to stratify therapeutic approaches for Ph+ leukemia patients.Genotyping And Cell Cycle

[0166] In addition to detecting the aforementioned cell states, the method may further comprise determining the genotype of the residual single cells. As established herein, mutation burden alone is insufficient when untethered from the cell state and the overall fitness that the cell state provides for the mutation to drive further disease progression. However, in some instances, knowledge of existing mutations may be used to help further select the appropriate therapeutic compound, for example, the specific combination of TKI inhibitor and SYK inhibitor or p38 inhibitor, or may further inform the need to supplement the treatment options disclosed herein with a further therapeutic compound of a different class, e.g., a kRas inhibitor.

[0167] Likewise, further assessing the residual cell’s cell cycle or development state may be helpful. Methods for identifying cell cycle states are known in the art. Methods for assessing developmental state, especially in leukemia, are described in the working examples provided herein.Treatment Options Based On Detected Cell State

[0168] As noted previously, the detected cell state may be used to select an appropriate therapeutic regimen for the subject suffering from cancer. In an embodiment, if a pre-BCR state is detected, either (i) an inhibitor of BCR-signaling is administered or (ii) a TKI inhibitor is administered in combination with a SYK inhibitor. Alternatively, if a stress / autophagy cell state is detected, a TKI inhibitor is administered in combination with the p38 inhibitor. In another embodiment, the subject may have completed induction therapy and proceeded to consolidation therapy. In an embodiment, the consolidation therapy may comprise a TKI inhibitor.

[0169] In an embodiment, the subject is in relapse following induction therapy. The indication therapy may comprise chemotherapy. Different combinations of chemo drugs mightbe used, but they typically include vincristine, dexamethasone, or prednisone, and an anthracycline drug such as doxorubicin (Adriamycin) or daunorubicin. In an embodiment, the induction therapy may comprise a TKI inhibitor. In another embodiment, the induction therapy may include chemotherapy in combination with a TKI inhibitor.Treatment Based On Detected Pre-BCR Cell State

[0170] In an embodiment, either (i) one or more BCR signaling pathway inhibitors or (ii) one or more TKI inhibitors in combination with one or more SYK inhibitors are administered, and a pre-BCR cell state is detected. In another embodiment, a first treatment comprising a BCR signaling pathway is administered followed by a second treatment comprising a TKI inhibitor with or without a SYK inhibitor.

[0171] As detailed in the working examples below, pre-BCR signaling and downstream ERK-mediated gene expression are associated with increased immature B marker gene expression in human patients. Accordingly, administering a BCR signaling pathway inhibitor may delay or prevent progression early in the treatment and before or in combination with further TKI therapy. In an embodiment, a BCR signaling pathway inhibitor may be an antibody, scFV, or nanobody capable of binding to the B-cell receptor. In another embodiment, the BCR signaling pathway inhibitor is an extracellular signal-regulated kinase (ERK) signaling pathway inhibitor. In an embodiment, the ERK signaling pathway inhibitor is trametinib, cobimetinib, selumetinib, ulixertinib (BVD-523), SCH772984, binimetinib, vemurafenib, dabrafenib, or combinations thereof.

[0172] Alternatively, the treatment may comprise a TKI inhibitor in combination with an SYK inhibitor. Tyrosine kinase inhibitors (TKIs) have emerged as a pivotal component in leukemia treatment. The advent of TKIs has significantly altered the therapeutic landscape of leukemia. TKIs are designed to target and inhibit the activity of specific tyrosine kinases that are abnormally active in cancer cells. SYK inhibitors target the spleen tyrosine kinase (SYK), a cytoplasmic tyrosine kinase that plays a critical role in the signaling pathways of the immune system. SYK is involved in various cellular processes, including proliferation, differentiation, and survival, particularly in B and myeloid cells. SYK is a critical player in B-cell receptor signaling, which is essential for developing functioning B cells. Inhibiting SYK can disrupt aberrant BCR signaling in these leukemic cells, impairing their growth and survival.In an embodiment, the TKI inhibitor may be deucravacitinib, avapritinib, capmatinib, pemigatinib, ripretinib, selpercatinib, selumetinib, tucatinib, entrectinib, erdafitinib, fedratinib,pexidartinib, upadacitinib, zanubrutinib, baricitinib, binimetinib, dacomitinib, fostamatinib, gilteritinib, Larotrectinib, lorlatinib, acalabrutinib, brigatinib, midostaurin, neratinib, alectinib, cobimetinib, Lenvatinib, Osimertinib, ceritinib, nintedanib, afatinib, ibrutinib, trametinib, axitinib, bosutinib, cabozantinib, ponatinib, regorafenib, tofacitinib, crizotinib, ruxolitinib, vandetanib, pazopanib, lapatinib, nilotinib, dasatinib, sunitinib, sorafenib, erlotinib, gefitinib, iormatinib. In an embodiment, the SYK inhibitor may be fostamiatinib (R788), entospletinib (GS-9973), TAK-659, or cerdulatinib (PRT0062070).Treatment Based On Detected Stress / Autophagy Cell State

[0173] In an embodiment, one or more TKI inhibitors and one or more p38 inhibitors are administered when a stress / autophagy cell state is detected. The one or more TKI inhibitors may be any TKI inhibitors described above. p38 inhibitors are a class of compounds that target the p38 mitogen-activated protein (MAP) kinases, which play crucial roles in cellular responses to stress and inflammation. These inhibitors can modulate key cell proliferation, apoptosis, and inflammation pathways. In an embodiment, one or more p38 inhibitors may be losmapimod, ruxolitinib, SB 2035280, doramapimod (BIRB 796), or panapimod (R-1503).Routes of Administration and Dosage Forms

[0174] The pharmaceutical formulations described herein can be administered to a subject in need thereof via any suitable method or route. Suitable administration routes can include, but are not limited to auricular (otic), buccal, conjunctival, cutaneous, dental, electro-osmosis, endocervical, endosinusial, endotracheal, enteral, epidural, extra-amniotic, extracorporeal, hemodialysis, infiltration, interstitial, intra-abdominal, intra-amniotic, intra-arterial, intraarticular, intrabiliary, intrabronchial, intrabursal, intracardiac, intracartilaginous, intracaudal, intracavernous, intracavitary, intracerebral, intracisternal, intracorneal, intracoronal (dental), intracoronary, intracorporus cavemosum, intradermal, intradiscal, intraductal, intraduodenal, intradural, intraepidermal, intraesophageal, intragastric, intragingival, intraileal, intralesional, intraluminal, intralymphatic, intramedullary, intrameningeal, intramuscular, intraocular, intraovarian, intrapericardial, intraperitoneal, intrapleural, intraprostatic, intrapulmonary, intrasinal, intraspinal, intrasynovial, intratendinous, intratesticular, intrathecal, intrathoracic, intratubular, intratumor, intratympanic, intrauterine, intravascular, intravenous, intravenous bolus, intravenous drip, intraventricular, intravesical, intravitreal, iontophoresis, irrigation, laryngeal, nasal, nasogastric, occlusive dressing technique, ophthalmic, oral, oropharyngeal, other, parenteral, percutaneous, periarticular, peridural, perineural, periodontal, rectal,respiratory (inhalation), retrobulbar, soft tissue, subarachnoid, subconjunctival, subcutaneous, sublingual, submucosal, topical, transdermal, transmucosal, transplacental, transtracheal, transtympanic, ureteral, urethral, and / or vaginal administration, and / or any combination of the above administration routes, which typically depends on the disease to be treated and / or the active ingredient(s).

[0175] In further embodiments, the amount of the primary active agent and / or optional secondary agent can be an effective amount, least effective amount, and / or therapeutically effective amount. As used herein, “effective amount,” “effective concentration,” and / or the like refer to the amount, concentration, etc., of the primary and / or optional secondary agent in the pharmaceutical formulation that achieves one or more therapeutic effects or desired effects. As used herein, “least effective,” “least effective concentration,” and / or the like amount refers to the lowest amount, concentration, etc., of the primary and / or optional secondary agent that achieves one or more therapeutic or other desired effects. As used herein, “therapeutically effective amount,” “therapeutically effective concentration,” and / or the like refer to the amount, concentration, etc., of the primary and / or optional secondary agent included in the pharmaceutical formulation that achieves one or more therapeutic effects. In further embodiments, one or more therapeutic effects are inducing an immune response in a subject to which they are delivered, inducing a B- and / or T-cell response in a subject to which it is delivered, treating or preventing a viral infection in a subject to which it is delivered.

[0176] The effective amount, least effective amount, and / or therapeutically effective amount of the primary and optional secondary active agent described elsewhere herein contained in the pharmaceutical formulation can be any non-zero amount ranging from about 0 to 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320, 330, 340, 350, 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470, 480, 490, 500, 510, 520, 530, 540, 550, 560, 570, 580, 590, 600, 610, 620, 630, 640, 650, 660, 670, 680, 690, 700, 710, 720, 730, 740, 750, 760, 770, 780, 790, 800, 810, 820, 830, 840, 850, 860, 870, 880, 890, 900, 910, 920, 930, 940, 950, 960, 970, 980, 990, 1000 pg, ng, pg, mg, or g or be any numerical value or subrange within any of these ranges.

[0177] In further embodiments, the effective amount, least effective amount, and / or therapeutically effective amount can be an effective concentration, least effective concentration, and / or therapeutically effective concentration, which can each be any non-zeroamount ranging from about 0 to 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320, 330, 340,350, 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470, 480, 490, 500, 510, 520, 530,540, 550, 560, 570, 580, 590, 600, 610, 620, 630, 640, 650, 660, 670, 680, 690, 700, 710, 720,730, 740, 750, 760, 770, 780, 790, 800, 810, 820, 830, 840, 850, 860, 870, 880, 890, 900, 910,920, 930, 940, 950, 960, 970, 980, 990, 1000 pM, nM, pM, mM, or M or be any numerical value or subrange within any of these ranges.

[0178] In other embodiments, the effective amount, least effective amount, and / or therapeutically effective amount of the primary and optional secondary active agent be any non-zero amount ranging from about 0 to 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320,330, 340, 350, 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470, 480, 490, 500, 510,520, 530, 540, 550, 560, 570, 580, 590, 600, 610, 620, 630, 640, 650, 660, 670, 680, 690, 700,710, 720, 730, 740, 750, 760, 770, 780, 790, 800, 810, 820, 830, 840, 850, 860, 870, 880, 890,900, 910, 920, 930, 940, 950, 960, 970, 980, 990, 1000 IU or be any numerical value or subrange within any of these ranges.

[0179] In further embodiments, the primary and / or the optional secondary active agent present in the pharmaceutical formulation can be any non-zero amount ranging from about 0 to 0.001, 0.002, 0.003, 0.004, 0.005, 0.006, 0.007, 0.008, 0.009, 0.01, 0.02, 0.03, 0.04, 0.05, 0.06, 0.07, 0.08, 0.09, 0.1, 0.11, 0.12, 0.13, 0.14, 0.15, 0.16, 0.17, 0.18, 0.19, 0.2, 0.21, 0.22, 0.23, 0.24, 0.25, 0.26, 0.27, 0.28, 0.29, 0.3, 0.31, 0.32, 0.33, 0.34, 0.35, 0.36, 0.37, 0.38, 0.39, 0.4, 0.41, 0.42, 0.43, 0.44, 0.45, 0.46, 0.47, 0.48, 0.49, 0.5, 0.51, 0.52, 0.53, 0.54, 0.55, 0.56, 0.57, 0.58, 0.59, 0.6, 0.61, 0.62, 0.63, 0.64, 0.65, 0.66, 0.67, 0.68, 0.69, 0.7, 0.71, 0.72, 0.73, 0.74, 0.75, 0.76, 0.77, 0.78, 0.79, 0.8, 0.81, 0.82, 0.83, 0.84, 0.85, 0.86, 0.87, 0.88, 0.89, 0.9, 0.91, 0.92, 0.93, 0.94, 0.95, 0.96, 0.97, 0.98, 0.9, to 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14,15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39,40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64,65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89,90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 99.1, 99.2, 99.3, 99.4, 99.5, 99.6, 99.7, 99.8, 99.9 % w / w, v / v, or w / v of the pharmaceutical formulation or be any numerical value or subrange within any of these ranges.

[0180] In further embodiments, the pharmaceutical formulations described herein can be provided in a dosage form. The dosage form can be administered to a subject in need thereof. The dosage form can effectively generate specific concentrations, such as an effective concentration, at a given site in the subject in need thereof. As used herein, “dose,” “unit dose,” or “dosage” can refer to physically discrete units suitable for use in a subject, each unit containing a predetermined quantity of the primary active agent and optionally present secondary active ingredient and / or a pharmaceutical formulation thereof calculated to produce the desired response or responses in association with its administration. In further embodiments, the given site is proximal to the administration site. In further embodiments, the given site is distal to the administration site. In some cases, the dosage form contains a greater amount of one or more of the active ingredients present in the pharmaceutical formulation than the final intended amount needed to reach a specific region or location within the subject to account for loss of the active components such as via first and second pass metabolism.

[0181] The dosage forms can be adapted for administration by any appropriate route. Appropriate routes include but are not limited to, oral (including buccal or sublingual), parenteral, subcutaneous, intramuscular, intravenous, and intradermal. Other appropriate routes are described elsewhere herein. Such formulations can be prepared using any method known in art.

[0182] Dosage forms adapted for oral administration can be discrete dosage units such as capsules, pellets or tablets, powders or granules, solutions, or suspensions in aqueous or nonaqueous liquids; edible foams or whips, or in oil-in-water liquid emulsions or water-in-oil liquid emulsions. In further embodiments, the pharmaceutical formulations adapted for oral administration include one or more agents that flavor, preserve, color, or help disperse the pharmaceutical formulation. Dosage forms prepared for oral administration can also be a liquid solution delivered as a foam, spray, or liquid solution. The oral dosage form can be administered to a subject in need thereof. Where appropriate, the dosage forms described herein can be microencapsulated.

[0183] The dosage form can also be prepared to prolong or sustain the release of any ingredient. In further embodiments, compounds, molecules, compositions, vectors, vector systems, cells, or a combination thereof described herein can be the ingredient whose release is delayed. In further embodiments, the primary active agent is the ingredient whose release is delayed. In additional embodiments, an optional secondary agent can be the ingredient whoserelease is delayed. Suitable methods for delaying the release of an ingredient include, but are not limited to, coating or embedding the ingredients in material in polymers, wax, gels, and the like. Delayed release dosage formulations can be prepared as described in standard references such as "Pharmaceutical dosage form tablets," eds. Liberman et al. (New York, Marcel Dekker, Inc., 1989), "Remington - The science and practice of pharmacy,” 20th ed., Lippincott Williams & Wilkins, Baltimore, MD, 2000, and "Pharmaceutical dosage forms and drug delivery systems,” 6th Edition, Ansel et al., (Media, PA: Williams and Wilkins, 1995). These references provide information on excipients, materials, equipment, and processes for preparing tablets and capsules and delayed-release dosage forms of tablets, pellets, capsules, and granules. The delayed release can be anywhere from about an hour to about three months or more.

[0184] Examples of suitable coating materials include but are not limited to, cellulose polymers such as cellulose acetate phthalate, hydroxypropyl cellulose, hydroxypropyl methylcellulose, hydroxypropyl methylcellulose phthalate, and hydroxypropyl methylcellulose acetate succinate; polyvinyl acetate phthalate, acrylic acid polymers and copolymers, and methacrylic resins that are commercially available under the trade name EUDRAGIT® (Roth Pharma, Westerstadt, Germany), zein, shellac, and polysaccharides.

[0185] Coatings may be formed with a different ratio of water-soluble polymer, waterinsoluble polymers, and / or pH-dependent polymers, with or without water-insoluble / water- soluble non-polymeric excipient, to produce the desired release profile. The coating is performed in the dosage form (matrix or simple), which includes. Still, it is not limited to tablets (compressed with or without coated beads), capsules (with or without coated beads), beads, particle compositions, and "ingredient as is" formulated as, but not limited to, suspension form or as a sprinkle dosage form.

[0186] Where appropriate, the dosage forms described herein can be a liposome. In these embodiments, primary active ingredient(s), and / or optional secondary active ingredient(s), and / or a pharmaceutically acceptable salt thereof, where appropriate, are incorporated into a liposome. In embodiments where the dosage form is a liposome, the pharmaceutical formulation is thus liposomal. The liposomal formulation can be administered to a subject in need thereof.

[0187] Dosage forms adapted for parenteral administration and / or adapted for inj ection can include aqueous and / or non-aqueous sterile injection solutions, which can contain antioxidants,buffers, bacteriostats, and solutes that render the composition isotonic with the blood of the subject, and aqueous and non-aqueous sterile suspensions, which can include suspending agents and thickening agents. The dosage forms adapted for parenteral administration can be presented in single-unit dose or multi-unit dose containers, including but not limited to sealed ampoules or vials. The doses can be lyophilized and re-suspended in a sterile carrier to reconstitute the dose before administration. Extemporaneous injection solutions and suspensions can be prepared from sterile powders, granules, and tablets in further embodiments. The parenteral formulations can be administered to a subject in need thereof.

[0188] For further embodiments, the dosage form contains a predetermined amount of a primary active agent, secondary active ingredient, and / or pharmaceutically acceptable salt thereof where appropriate per unit dose. In an embodiment, where applicable, the predetermined amount of primary active agent, secondary active ingredient, and / or pharmaceutically acceptable salt can be an effective, least effective, and / or therapeutically effective amount. In other embodiments, the predetermined amount of a primary active agent, secondary active agent, and / or pharmaceutically acceptable salt thereof, where appropriate, can be a proper fraction of the effective amount of the active ingredient.

[0189] The pharmaceutical formulations or dosage forms thereof described herein can be administered one or more times hourly, daily, monthly, or yearly (e.g., 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or more times hourly, daily, monthly, or yearly). In further embodiments, the pharmaceutical formulations or dosage forms described herein can be administered continuously over a period ranging from minutes to hours to days. Devices and dosage forms are known in the art and described herein that are effective in providing continuous administration of the pharmaceutical formulations described herein. In further embodiments, the first or a few initial amounts (s) administered can be higher than subsequent doses. This is typically referred to in the art as a loading dose or doses and a maintenance dose, respectively. In further embodiments, the pharmaceutical formulations can be administered such that the doses are tapered (increased or decreased) over time to wean a subject gradually off of a pharmaceutical formulation or introduce progressively a subject to the pharmaceutical formulation.

[0190] As previously discussed, the pharmaceutical formulation can contain a predetermined amount of a primary active agent, secondary active agent, and / or pharmaceutically acceptable salt thereof where appropriate. In an embodiment, thepredetermined amount can be an appropriate fraction of the effective amount of the active ingredient. Such unit doses may, therefore, be administered once or more than once a day, month, oryear (e.g., 1, 2, 3, 4, 5, 6, or more times per day, month, oryear). Such pharmaceutical formulations may be prepared using any of the methods well-known in the art.

[0191] Where co-therapies or multiple pharmaceutical formulations are to be delivered to a subject, the different therapies or formulations can be administered sequentially or simultaneously. Sequential administration is when an appreciable amount of time occurs between administrations, such as more than 15, 20, 30, 45, 60 minutes or more. The time between administrations in sequential administration can be on the order of hours, days, months, or even years, depending on the active agent present in each administration. Simultaneous administration refers to the administration of two or more formulations at the same time or substantially at the same time (e.g., within seconds or just a few minutes apart), where the intent is that the formulations be administered together at the same time.

[0192] The following examples illustrate further embodiments. They are given for illustrative purposes only and are not intended to limit the scope of the invention.EXAMPLESExample 1 - EXPERIMENTAL MODEL AND SUBJECT DETAILSGeneration and Use of PDXs

[0193] Primary bone marrow and peripheral blood specimens were collected from patients with leukemia at the Dana-Farber Cancer Institute, Brigham and Women’s Hospital, and Boston Children’s Hospital for xenotransplantation. Additional PDXs already established through the Public Repository of Xenografts (PRoXe) were utilized.1De-identified patient samples were obtained with informed consent and xenografted under Dana-Farber / Harvard Cancer Center Institutional Review Board (IRB)-approved protocols #13-351 and #17-348. ^o< .C -PrkdcscldIL2r^mIW}lI zi (NSG) mice were purchased from Jackson Laboratories and handled according to Dana-Farber Cancer Institute's Institutional Animal Care and Use Committee-approved protocols #13-034 and #18-021.In vivo therapeutic studies

[0194] Viable frozen B-ALL xenograft cells were thawed and changed into 1 x PBS before tail-vein injection at 0.5-2.0 x 106cells per mouse. Engraftment was monitored by weekly peripheral blood (50pL) flow cytometry beginning three weeks after injection. Blood wasprocessed with Red Blood Cell Lysis Buffer (Qiagen #158904; Hilden, Germany) and stained with antibodies against human CD45 (APC-conjugated, eBioscience #17-0459-42; San Diego, CA, USA) and human CD19 (PE-conjugated, eBioscience #12-0193-82) in 1 x PBS with EDTA (2mM). Flow cytometry data were analyzed using Flow Jo software (BD Biosciences; Ashland, OR, USA). Upon engraftment-when>10% of cells were positive for CD45 and CD19-mice within each PDX line underwent 1 :2:2:4: 1 randomization to the following arms and initiated treatment within two days: (1) sacrifice for baseline tissue interrogation; (2) ponatinib (Selleckchem #S1490; Houston, TX, USA; constituted in 25 mM citrate buffer, pH 2.75) 40mg / kg via oral gavage (OG) daily; (3) NVP-ABL001 (Novartis Pharmaceuticals; Basel, Switzerland; constituted in HC1 0.1 M, PEG300 30%, Solutol HS15 6%, NaOH IM, sodium acetate buffer pH 4.7 lOmM) 30 mg / kg OG twice daily; (4) ponatinib 40 mg / kg OG twice daily plus NVP-ABL001 30 mg / kg OGB ID; and (5) vehicle (alternating doses of vehicle used for ponatinib and NVP-ABL001, at equivalent volumes). One mouse per active treatment arm per PDX line was sacrificed for pharmacodynamic assessment on day 7 of treatment. The remaining mice continued daily treatment under monitoring with biweekly peripheral blood flow (lOOpL) cytometry until progression (defined as peripheral blood involvement of at least 10% on two consecutive assessments at least one week apart), weight loss of greater than 20% from pre-treatment baseline, or clinical manifestations of advanced disease, including but not limited to ruffled fur, hunched posture, hind limb paralysis, or lethargy. Progression or toxicity defined above triggered humane euthanasia by CO2 asphyxiation, necropsy to ascertain or corroborate the cause of death, and post-mortem harvest of peripheral blood, bone marrow, and any apparent soft tissue masses. Additional in vivo studies involved treatment with nilotinib (Selleckchem #S1033), which was constituted inN-methyl-2-pyrrolidone (10%) in polyethylene glycol (PEG)-300 (90%) and dosed at 50mg / kg OG twice daily.Human donors for reference (immunophenotyping, scRNA-seq)

[0195] Normal human bone marrow aspirates were obtained from donors who had provided written informed consent for tissue banking for research purposes under Dana- Farber / Harvard Cancer Center protocol #01-206 and were undergoing bone marrow harvest for unrelated hematopoietic stem cell transplantation recipients. Briefly, bone marrow was collected into a Baxter harvest collection system with diluent consistent sodium heparin in lactated Ringers solution at 80 units / mL. Bone marrow was heparinized at a concentration of 15,000-20,000 units / mL and filtered inline using 200- and 500-micron filters. Bone marrowmononuclear cells from the heparinized, filtered product were isolated via density gradient centrifugation (Ficoll-Paque, ThermoFisher Scientific #45-001-749). Lymphoid progenitors were enriched by fluorescence-activated cell sorting (FACS) based on staining with the following cell surface markers: CD45, CD34, CD19, CD10, CD20, and CD22. Lymphoid progenitor sub-populations then underwent single-cell RNA sequencing.Phase I clinical trial

[0196] Serial primary blood and bone marrow specimens were obtained from appropriately consented patients treated on a phase I, investigator-initiated clinical trial (NCT03595917) of asciminib (ABL001) in combination with dasatinib plus prednisone for adults with newly diagnosed BCR-ABL+ ALL or chronic myelogenous leukemia in lymphoid blast crisis (CML- LBC). These patients cross-consented to Dana-Farber Cancer Institute tissue banking protocol #01-206. Bone marrow was obtained at screening and after each 21 -day cycle through the first four cycles. Peripheral blood was obtained at screening and on days 2, 4, 8, 11, 15, and 22 (±2 days) of cycle 1. Both bone marrow and peripheral blood were collected into EDTA vacutainer tubes before mononuclear cell isolation per standard protocols. Bone marrow and peripheral blood underwent clinical quantitative real-time PCR for BCR-ABL1 mRNA according to the BCR-ABL isoform detected at screening (pl 90 or p210). Curated sets of BCR- AB L+ ALL clinically annotated specimens underwent evaluation by single-cell RNA-sequencing ± biophysical profiling as part of the trial’s correlative science program.Example 2 - METHOD DETAILSQuantifying BCR-ABL1 mRNA in PDX peripheral blood with qRT-PCR

[0197] BCR-ABL mRNA levels were measured via quantitative real-time PCR (qRT- PCR) of serial peripheral blood specimens from PDX models to track the kinetics of response and progression. Briefly, xenografted mice were phlebotomized for 100 pL by submandibular vein laceration every two weeks. Blood was stored in RNAProtect tubes (Qiagen #76544). Messenger RNA was isolated using the RNeasy Protect Animal Blood Kit (Qiagen #73224) and quantified using the iScript One-Step RT-PCR Kit with SYBR Green (Bio-Rad #170- 8893) on a Bio-Rad CFX96 Thermal Cycler. Synthesis of cDNAs was performed with random hexamers. Amplification of cDNAs was performed using iTaq Universal SYBR Green Supermix (Bio-Rad #172-5125) and the following oligomers:BCR-ABL isoform pl90 forward: CAACAGTCCTTCGACAGCAG (SEQ ID NO: 1) BCR-ABL isoform pl90 reverse: CCCTGAGGCTCAAAGTCAGA (SEQ ID NO: 2)BCR-ABL isoform p210 forward: TCCGCTGACCATCAATAAGGA (SEQ ID NO: 3) BCR-ABL isoform p210 reverse: CACTCAGACCCTGAGGCTCAA (SEQ ID NO: 4)

[0198] Positive control reagents for each isoform were pl 90 clonal control RNA (Invivoscribe #4-089-2800) and mRNA isolated from the BCR-ABL p210-positive cell line K562. BCR-ABL mRNA was quantified in the peripheral blood of patients treated on clinical trial protocol #18-170 via CAP / CLIA-approved clinical BCR-ABL qRT-PCR performed in the clinical molecular laboratory of Brigham and Women’s Hospital.Quantifying BCR-ABL1 mRNA in primary patient peripheral blood with qRT-PCR

[0199] BCR-ABL mRNA was quantified in the peripheral blood of patients treated on clinical trial protocol #18-170 via CAP / CLIA-approved clinical BCR-ABL qRT-PCR performed in the clinical molecular laboratory of Brigham and Women’s Hospital.Targeted DNA Sequencing

[0200] PDX models underwent mutational profiling with targeted panels. Leukemia cells were enriched from fresh primary PDX bone marrow or peripheral blood via immunomagnetic enrichment for human B cells using human CD19 MicroBeads (Miltenyi Biotec#130-050-301; Gaithersburg, MD, USA). DNA was extracted using the DNeasy Blood & Tissue kit (Qiagen #69504) and fluorometrically quantitated using the Qubit dsDNA HS assay kit (Invitrogen #Q32854; Waltham, MA, USA) before use in next-generation sequencing library preparation.

[0201] A hybrid-capture target enrichment panel targeting the full coding sequences of 183 genes selected based on the presence of recurrent mutations in hematologic malignancies was utilized to profile most PDX models at baseline, on-treatment, and at the end of the study (as previously described with panel details in Table S2). An amplicon-based clinical sequencing panel targeting hotspot regions of the oncogenes and most of the coding regions of tumor suppressor genes recurrently implicated in hematologic malignancies was employed for a subset of PDX models (total 93 genes). A custom amplicon-based deep sequencing panel targeting 23 genes implicated in B-ALL treatment resistance (ArcherDX; Boulder, CO, USA) was employed to profile PDXs progressing after BCR-ABL inhibition.Whole Exome Sequencing (WES) sample preparation

[0202] PDXs that progressed without treatment-emergent driver alterations detected by targeted sequencing underwent whole exome sequencing using the SureSelect Human All Exon v5 kit (Agilent Life Sciences; Santa Clara, CA, USA). Briefly, 100 ng of genomic DNA from each leukemia specimen, as well as a control cell line (CEPH 1408) and a tail clippingfrom a non-xenografted NSG mouse, were fragmented to 250 bp on a Covaris Ultrasonicator (Woburn, MA, USA). Size-selected DNA fragments were ligated to xGen vl UDI-UMI9 adaptors (Integrated DNA Technologies; Coralville, IA, USA) during automated library preparation with a Biomek FXPliquid handling robot (Beckman Coulter; Indianapolis, IN, USA). Libraries (250 ng per sample) were pooled to 750 ng and captured with the SureSelect Human All Exon v5 bait set. Captures were pooled and sequenced on a HiSeq 3000 (Illumina; San Diego, CA, USA).Flow sorting from healthy human bone marrow aspirates and PDX tumors.

[0203] Approximately 106cells per sample were resuspended in PBS with 4,6-diamidino- 2-phenylindole (DAPI; 0.75 pg ml-1) as a dead-cell marker. For cell surface staining, PBS- washed cells were blocked with an Fc blocker for 10 min on ice and stained with the indicated antibodies (listed in Supplementary Table 11) or with an isotype control for 25 min on ice. Cells were washed and resuspended in chilled PBS containing 0.75 pg ml-1 of DAPI to exclude dead cells. For annexin V staining, annexin V binding buffer (BD Bioscience) was used instead of PBS, and 7-aminoactinomysin D (7-AAD; BD Bioscience) was used instead of DAPI. Phycoerythrin (PE)-labeled annexin V, was purchased from BD Bioscience. The acquisition was performed on an LSRFortessa flow cytometer (BD Biosciences). Fluorescence-based cell sorting was performed on a FACSAria II (BD Biosciences). FACS data were analysed using FlowJo software (FlowJo).

[0204] Flow-sorted gates for B cell progenitor populations slightly differed between healthy human cells and leukemic PDX cells, as outlined in Figure 5A-5B and Figure 6A. Given the acquisition of CD22 in CD34-CD10-CD19+ cells before CD20 in leukemic cells, Pre-B cells were defined as CD34-CD10-CD19+CD20-CD22+ compared to their canonical definition as CD34-CD10-CD19+CD20+CD22-.Sample preparation for scRNA-sequencing of clinical and PDX samples

[0205] Applicants used the Seq-Well S3platform for massively parallel scRNA-seq to capture transcriptomes of single cells on barcoded mRNA capture beads. Briefly, a single-cell suspension of 15,000 cells in 200pL RPMI media supplemented with 10% FBS was loaded onto single arrays containing barcoded mRNA capture beads (ChemGenes). The arrays were sealed with a polycarbonate membrane (pore size of 0.01 pm) before cell lysis and transcript hybridization. The barcoded mRNA capture beads were then recovered and pooled for all subsequent steps. Reverse transcription was performed using Maxima H Minus ReverseTranscriptase (Thermo Fisher Scientific EP0753). Exonuclease I treatment (NEB M0293 L) removed excess primers. Second-strand synthesis used a primer of eight random bases to create complementary cDNA strands with SMART handles for PCR amplification. Whole transcriptome amplification was done using KAPA HiFi PCR Mastermix (Kapa Biosystems KK2602) with 2000 beads per 50-pl reaction volume. Libraries were then pooled in sets of eight (totaling 16,000 beads), purified using Agencourt AMPure XP beads (Beckman Coulter, A63881) by a 0.6* solid phase reversible immobilization (SPRI) followed by a 0.8* SPRI, and quantified using Qubit hsDNA Assay (Thermo Fisher Scientific Q32854). The quality of the whole transcriptome amplification (WTA) product was assessed using the Agilent High Sensitivity D5000 Screen Tape System (Agilent Genomics) with an expected peak at 800 base pairs tailing off to beyond 3000 base pairs and a small / nonexistent primer peak.

[0206] Libraries were constructed using the Nextera XT DNA tagmentation method (Illumina FC-131-1096) on a total of 750 pg of pooled cDNA library from 16,000 recovered beads using index primers with format as previously described. Tagmented and amplified sequences were purified at a 0.8* SPRI ratio, yielding library sizes with an average distribution of 500 to 750 bp in length as determined using the Agilent High Sensitivity DI 000 Screen Tape System (Agilent Genomics). Two arrays were sequenced per sequencing run with an Illumina 75 Cycle NextSeq500 / 550v2 kit (Illumina FC-404-2005) at a final concentration of 2.4 pM. The read structure was paired with Read 1 starting from a custom Read one primer containing 20 bases with a 12-bp cell barcode and 8-bp unique molecular identifier (UMI) and Read 2 containing 50 bases of transcript sequence.Sample preparation for paired SMR mass profiling and SMART-Seq2

[0207] For all PDX and healthy bone marrow samples, cells were adjusted to a final concentration of 2.5 x 105cells / ml to load single cells into the mass sensor array and record single-cell mass measurements, as previously described. In order to exchange buffer and flush individual cells from the system, the release side of the device was constantly flushed with PBS at a rate of 15 pL per minute. Upon detection of a single cell at the final cantilever of the sSMR, as indicated by a supra-threshold shift in resonant frequency, a set of three-dimensional motorized stages (Thorlabs) was triggered to move a custom PCR-tube strip mount from a waste collection position to a sample collection position to retrieve the cell. Each cell was dispensed in approximately five pl of PBS into a PCR tube containing five pl of 2 TCL lysis buffer (Qiagen) with 2% vlv 2-mercaptoethanol (Sigma) for a total final reaction volume of10 pl. After each 8-tube PCR strip was filled with cells, the strip was spun down at 1000 g for 30 s and immediately snap-frozen on dry ice. Following collection, samples were stored at - 80°C before library preparation and sequencing.

[0208] Single-cell lysates were compiled from independent collections upon thawing and transferred into wells of a 0.2mL skirted 96-well PCR plate (Thermo Fisher Scientific). Singlecell RNA-sequencing (scRNA-seq) libraries were generated using the SMART-Seq2 protocol. Briefly, cDNA was reversed and transcribed from single cells using Maxima RT (Thermo Fisher Scientific), and whole transcriptome amplification (WTA) was performed. WTA products were purified using the Agencourt AMPure XP beads (Beckman Coulter), which were used to prepare paired-end libraries with Nextera XT (Illumina). Single cells were pooled and sequenced on a NextSeq 550 sequencer (Illumina) using a 75-cycle High Output Kit (v2.5) with a 30-bp paired-end read structure.Example 3 - QUANTIFICATION AND STATISTICAL ANALYSISPDX in vivo studies: Defining pretreatment clinical risk scorePDX in vivo studies: survival analysis on treatment arms and with pretreatment clinical risk stratification metadata

[0209] Analyses fitting a Cox proportional hazards model for overall survival (OS) and progression-free survival (PFS) outcomes on treatment arms and pretreatment clinical risk stratification categories were performed using the survival package in R. Hazard ratios and p- values for PFS within pretreatment clinical risk categories were generated relative to the lowest risk group in each category. The overall preclinical risk score was generated from a weighted sum of each categorical risk group outlined in Figure 2C.WES alignment and variant calling

[0210] Pooled sequenced WES samples were demultiplexed using Picard tools. Read pairs were aligned to the hgl9 reference build using the Burrows- Wheel er Aligner.4Data were sorted and duplicate-marked using Picard tools. Alignments were refined using the Genome Analysis Toolkit (GATK)5’6for localized realignment around small insertion and deletion (indel) sites. Mutation analysis for single nucleotide variants was performed with MuTect vl.1.47and annotated by Variant Effect Predictor.8Indels were called using the SomaticIndelDetector tool of the GATK. Copy number variants (CNVs) were identified using RobustCNV for autosomes.scRNA-seq sequencing alignment and quality control

[0211] Sequenced Seq-Well BCL files were demultiplexed into individual sample FASTQs for Read one and Read 2 using the bcl2fastq pipeline on Terra, as previously described. The resultant paired read FASTQs were aligned to the hgl9 genome using the cumulus / dropseq tools pipeline on Terra maintained by the Broad Institute using standard settings, generating a genes by cells count matrix for each sample. Low-quality cells were filtered using nGene <= 200, nUMI <= 500, and percent mitochondrial transcripts <= 30% thresholds before merging samples; genes were filtered if they were not expressed in at least ten cells.

[0212] Sequenced SS2 BCL files were similarly demultiplexed using bcl2fastq and aligned to the hgl9 genome using publicly available scripts on Terra (github.com / broadinstitute / TAG- public). Total gene counts and transcript per million (TPM) matrices were filtered to remove low-quality cells with <15% transcriptome mapping, 2,000 genes, and 45,000 mapped reads before continuing analysis. Genes expressed in fewer than ten cells, long non-coding RNAs, and unique hgl9 reference-build variants were removed before downstream analysis.Human healthy bone marrow reference cell type clustering and visualization

[0213] After QC filtering, 13,643 high-quality cells from n=8 healthy human bone marrow donors were analyzed in Seurat v2.3.4 to classify hematopoietic cell types. After normalization, the top 1,500 highly dispersed variable genes were selected using the mean-variance plot method in Seurat’s FindVariableFeatures function. scRNA-seq data was scaled over highly variable genes and used as input for PCA analysis. The top significant PCs, as defined by the JackStraw test (top 25 PCs), were used as input for building an SNN graph to cluster cells by their (k=35) nearest neighbors and for t-SNE visualization of clusters. Given the shared, continuous hierarchy of covarying gene expression in hematopoietic development, broad cell types (progenitor, myeloid, erythroid, B cell lineage, pDCs, T cells, and Plasmablasts) were called based on their differentially expressed genes (identified using the Wilcox test in Seurat’s Find AllMarkers function), and subset into individual Seurat objects for a second round of clustering to resolve the final 13 cell types defined in Figure 6. Cell type annotations were post-hoc validated based on biased or exclusive expression of known marker genes (Figure 6D)

[0214] SS2 healthy reference cell types were called by their confident random forest prediction probabilities (see next section) and examination of marker genes further to supportcell type identification (Figure 6D-6F). Cell type clusters were visualized using SPRING, a tool that generates force-directed layouts from kNN graphs to preserve hierarchical relationships between cell types visually.Training and interpreting the random forest classifier

[0215] Random forest is an ensemble machine-learning method used for both classification and regression. Like other ensemble models, random forests combine multiple weak classifiers, in this case shallow decision trees, to make predictions. In this work, a random forest was used for classification. Here, Applicants interrogate aberrant developmental hierarchies in ALL using random forests to predict the nearest cell type from the typical B-cell lineage for single cells from B-ALL samples. There are inherent advantages to random forests for the B-ALL classification task. Importantly, ensemble classifiers, like a random forest, provide a distribution of class probabilities reflecting the similarity of each cell to each cell type the model was trained on. This is done by calculating the proportion of trees voting for a cell type for each observation. To generate a single prediction for a cell, the highest-class probability becomes the prediction. The higher the probability of the chosen class, the more transcriptionally similar the cell is to that stage of B-cell development. The distribution of class probabilities can be used to understand a prediction's certainty - or uncertainty -. Applicants leveraged this measure of uncertainty in predictions to evaluate how well a tumor cell fits a specific stage in the B-cell lineage (Figure 4F). A tumor cell with a more uniform distribution of probabilities over classes likely shares transcriptional features with many a wider range of stages of B-cell development, potentially indicating a more aberrant cell from normal development. Second, ensemble approaches tend to be more robust to overfitting, which is necessary when applying a model trained on sorted, healthy populations of cells to evaluate aberrant leukemic cells. Finally, because random forests are nonparametric models, they are also highly flexible to input feature scale and variance. These approaches are particularly suited to raw count matrices output by various scRNA-seq technologies.

[0216] Here, Applicants trained a random forest on [sorted cells from the B-cell lineage] using -15,000 genes as input features. Random forests were implemented using R version 3.5.1 using the caret package for training infrastructure1. The ranger implementation of random forests was used2. Hyperparameter search over ranger parameters (the number of randomly selected features considered for splitting at each tree node and the rule used for splitting) was done via 10-fold cross-validation (CV). The model achieved an accuracy of 94±0.006% on a10-fold CV with optimal parameters. The final model used the full training set of 13,643 cells. Results of 10-fold cross-validation are provided in Figure 7A. The model was also evaluated on an external testing set of Seq-Well-generated healthy bone marrow scRNA-seq transcriptomes and achieved a performance of average AUC=0.99 over all 13 cell types (Figure 7C). Applicants used permutation importance tests to interpret features being used to make predictions by the classifier. Permutation importance measures the impact of randomly shuffling feature values on the performance of a model, which is measured as accuracy and decreases Gini impurity. Specifically, a computationally accelerated heuristic method was used to construct a null distribution from features with importance values close to zero, limiting the need for randomly shuffling all features independently to evaluate significance3. The results of feature importance defining marker genes segregating the 13 cell types can be found in Figure 7BGenerating Tumor Hybrid Scores and assigning leukemic cells to hybrid populations

[0217] Tumor Hybrid gene signatures were generated as previously described. First, normalized gene expression values were correlated to RF cell type classification probabilities along B cell progenitor cell types (HSC, Pre-B, and Immature B). Pro-B RF probability correlations were excluded; since most leukemic cells were dominantly classified as Pro-B with secondary classifications along B cell lineage cell types, genes highly correlated to Pro-B RF probabilities were not Pro-B -specific. To ensure that genes in each hybrid population signature were specific and unique to HSC, Pre-B, and Immature B cell types, the second- highest cell type correlation coefficient was subtracted from the highest correlation coefficient for a given cell type. Additionally, to ensure that cell type signatures were not obfuscated by cell cycle, positive correlation values of genes with cell cycle scores were subtracted from the highest correlation coefficient of a given cell type. After performing these corrections, the top 30 correlated genes to HSC, Pre-B, and Immature B cell types were included in their respective hybrid gene signatures. Pro-B gene scores were defined by healthy Pro-B cells' top 30 differentially expressed genes.

[0218] Tumor cells were scored by these HSC, Pro-B, Pre-B, and Immature B gene signatures and consequently assigned to hybrid populations similarly to what has previously been described. Single-cells were classified into HSC-like, Pre-B-like, and Immature-like hybrid populations based on their highest hybrid cell type signature score, which needed 0.5 standard deviations above the mean Pro-B score. All other cells were classified as Pro-BHybrids, characterized by robust Pro B gene expression and weak or no co-expression of different cell type hybrid signature scores. The classifications based on these hybrid score distributions and relative to their B cell lineage RF prediction probabilities are demonstrated in Figure 8A.Mutual information of transcription factor activities with tumor hybrids

[0219] Applicants were interested in defining gene programs whose activity is associated with the tumor hybrid populations described above. Given the highly entropic co-expression of tumor hybrid signatures with Pro-B marker genes, Applicants utilized mutual information to measure the potentially non-linear mutual dependence of gene expression with hybrid- defined developmental marker genes. Within respective hybrid subpopulations of each PDX line’s pre-treatment and progression time points, Applicants calculated the average normalized mutual information (NMI) of all highly expressed genes across the top 30 genes in each hybrid population signature. Within each PDX sample and hybrid population, MI values between each gene-gene pair were generated using R info the package misinformation function with the Miller-Madow asymptotic bias corrected empirical estimator and normalized to scale values between 0 and 1 as a relative, comparable metric between samples. Applicants interpret these NMI values as a metric for genes whose expression is relatively scaled with hybrid population identity.

[0220] To identify cooperatively expressed genes that are collectively mutually informed with tumor hybrid signatures, Applicants utilized the collectRI transcription factor accessibility database along with the decoupleR package to in silico predict mutually informed transcription factor (TF) activity with tumor hybrids. Averaged NMI values for each PDX sample hybrid were used as input with the run ulm function to estimate the linear relationship between TF- target genes and their hybrid marker gene expression. Within each PDX sample, significant TFs were ordered by their variance in mutually informed activity between hybrid populations, and the top 30 of these TFs were selected for validation. Ward distance clustering and Pearson co-correlation of predicted mutually informed TF activities revealed three subclusters of TFs corresponding to HSC-like, Pre B-like, and Immature B-like hybrid populations (Figure 4J, Figure 8B). Applicants projected TF module scores over leukemic secondary RF prediction probabilities to validate the activity of these TFs associated with developmental marker gene expression.Defining developmental skews in Smart-seq2 PDX samples

[0221] Given the lack of RF-classified immature B cells in the SS2 leukemic dataset, Applicants identified Pearson genes correlated with Pre-B RF probabilities and progenitor population (HSC, GMP, Pro-Mono, Early-Erythroid) probabilities. Applicants found that genes associated with progenitor RF probabilities negatively correlated with Pre-B RF probabilities in leukemic cells and vice versa, enabling Applicants to define a spectrum of differentiation between progenitor and later-stage B cell developmental stages. Progenitor-like and Pre-B-like scores were generated by scoring leukemic cells over the top 30 genes significantly correlated to their respective RF probabilities. Each cell’s location on the leukemic differentiation spectrum was defined by its (Pre-B-like score - Progenitor-like score).Identifying somatic variants in full-length Smart-seq2 scRNA-seq libraries

[0222] Variants were called similar to previously described methods. Briefly, each sample's FASTQ files were aligned to hgl9 using STAR (version 2.6.0c) and then sorted and indexed with SAMtools (version 1.13). Then, 20 loci of interest (that were detected in bulk sequencing data) were assessed by a custom script utilizing the Pysam library (version 0.16.0.1). In particular, for each locus of interest, each sample was marked as "NC" if there was no coverage at the locus, marked with 0 if all overlapping reads matched the REF allele, or marked with the number of alternate reads if there were overlapping reads that did not match the REF allele.Predicting chromosomal number variations (CNVs) in SS2 scRNA-seq libraries with inferCNV

[0223] To identify SS2 leukemic cells harboring CNVs and in silico elucidate subclonal heterogeneity within tumors, applicants estimated single-cell CNVs as previously described by computing the average expression in a sliding window of 100 genes within each chromosome after sorting the detected genes by their hgl9 genome-defined chromosomal coordinates. Applicants used all healthy bone marrow SS2 cells identified above (Figure 7B) as reference normal populations for this analysis. Complete information on the inferCNV workflow used for this analysis can be found here: github.com / broadinstitute / inferCNV / wiki, using baseline input parameters for SS2 data and for the i6 HMM algorithm for confident CNV-positive or pessimistic predictions in single-cells.Module Scoring single-cell transcriptomes

[0224] Module scores of all gene signatures over single-cells were annotated using the Seurat v4 AddModule Score function, which calculates the average expression levels of genesin a gene list relative to all other genes with comparable normalized gene expression. ERK signaling module scores (Figure 9G, Figure 10B) were generated from the msigDB BIOCARTA ERK PATHWAY gene set. Quiescent cells were binned based on positive scores for xx. Applicants utilized previously established signatures for Gl / S (n = 43 genes) and G2 / M (n = 55 genes) to place each cell along this dynamic process; after inspecting the distribution of scores in the complete dataset, Applicants considered any cell > 1.5 SD above the mean for either the Gl / S or the G2 / M scores to be cycling. Senescence scores were derived from genes significantly differentially expressed in the SS2 DFAB25157 RAS-mutant cells in remission compared to all other RAS-mutant SS2 cells (Figure 11F).Defining stress-autophagy, pre-BCR signaling, and inflammation transcriptional programs in remission

[0225] To define heterogeneous, correlated transcriptional states defining PDX tumors that emerge in remission, Applicants first performed differential gene expression analysis between paired pre-treatment and remission cells within the same PDX line to identify genes that significantly increase expression at remission. A set of 40 remission state-defining genes was identified based on significant upregulation in at least two PDX-specific remission differentially expressed gene (DEG) lists. Performing gene-gene Pearson correlation across the expression of these 40 shared remission-high DEGs in all remission leukemic cells revealed three correlated modules of genes. To expand these three modules, Applicants identified genes significantly correlated (> two standard deviations above median Pearson correlation) with the top differentially expressed gene in each module. Pathway enrichment of significantly correlated genes was performed over msigDB Reactome gene sets for functional annotation and to nominate targeted inhibitors of state (Figure 17E, Figure 16D).Example 4 - Ph+ CELL LINE VALIDATION OF DEVELOPMENTAL STATE-DRUG RESPONSE RELATIONSHIP

[0226] Three Ph+ ALL (Philadelphia chromosome-positive acute lymphoblastic leukemia) cell lines (Z-181, Z-119, SUP-B15) were maintained in RPMI-1640 with 10% FBS. Singlecell RNA sequencing was performed using Seq-Well S3platform as described in Example 2. UMAP visualization revealed distinct transcriptional profiles (Figure 18A).Drug sensitivity analysis

[0227] Cells were treated with serial dilutions of dasatinib and ponatinib (0.1 nM to 1 pM) for 72 hours. Cell viability was measured by MTT assay. IC50 values were calculated using four-parameter logistic regression (Figure 18B).Random forest classification

[0228] Cell lines were classified using the trained random forest classifier from Example 3. Classification probabilities across B-cell developmental stages showed progressively more mature signatures: Z-181 (most progenitor-like) —> Z-l 19 —> SUP-B15 (most mature) (Figure 18C)Developmental maturity correlates with ABL inhibitor resistance

[0229] Cell lines with higher mature B-cell fractions demonstrated significantly higher IC50 values for both dasatinib and ponatinib (Figure 18D). SUP-B15 (most mature) showed greatest resistance, while Z-181 (most progenitor-like) showed highest sensitivity. This validates that developmental cell state determines therapeutic response in Ph+ ALL, supporting the predictive utility of developmental staging for ABL inhibitor sensitivity.Example 5 - MUTATION AND CELL STATE COMPATIBILITY IS REQUIRED AND TARGETABLE IN PH+ ACUTE LYMPHOBLASTIC LEUKEMIA MINIMAL RESIDUAL DISEASE

[0230] Efforts to cure BCR::ABL1 B cell acute lymphoblastic leukemia (Ph+ ALL) solely through inhibition of ABL1 kinase activity have thus far been insufficient despite the availability of tyrosine kinase inhibitors (TKIs) with broad activity against resistance mutants. The mechanisms that drive persistence within minimal residual disease (MRD) remain poorly understood and therefore untargeted. Utilizing 13 patient-derived xenograft (PDX) models and clinical trial specimens of Ph+ ALL, Applicants examined how genetic and transcriptional features co-evolve to drive progression during prolonged TKI response. Our work reveals a landscape of cooperative mutational and transcriptional escape mechanisms that differ from those causing resistance to first generation TKIs. By analyzing MRD during remission, Applicants show that the same resistance mutation can either increase or decrease cellular fitness depending on transcriptional state. Applicants further demonstrate that directly targeting transcriptional state-associated vulnerabilities at MRD can overcome BCR:: ABL 1 independence, suggesting a new paradigm for rationally eradicating MRD prior to relapse. Finally, Applicants illustrate how cell mass measurements of leukemia cells can be used torapidly monitor dominant transcriptional features of Ph+ ALL to help rationally guide therapeutic selection from low-input samples.

[0231] A large fraction of patients with cancer achieve complete remission at some point during their course of therapy, either through surgery, chemotherapy, radiation, or a combination thereof. Nevertheless, many of these patients relapse or progress owing to a small pool of remaining cancer cells commonly referred to as minimal residual disease (MRD). This is even true for cancers with clear, targetable oncogene dependencies such as BCR::ABL1- rearranged B cell acute lymphoblastic leukemia (Ph+ ALL). Despite highly effective tyrosine kinase inhibitors (TKI) with potent activity against multiple resistance-conferring point mutations in BCR::ABL1, relapse during single-agent treatment is nearly universal1’2’3. Unfortunately, accumulating evidence casts doubt on the potential for up-front combinations of next-generation TKIs to fully overcome subclonal heterogeneity and thereby eradicate MRD4

[0232] While most patients with BCR: :ABL1 -driven disease relapse with kinase domain mutations, 30-40% of patients progress with BCR: :ABL1 -independent mechanisms that are poorly understood5. Previous studies have identified developmental heterogeneity across ALL6’7, as well as in Ph+ ALL specifically8’9. This developmental heterogeneity has also been linked to treatment response for multiple classes of inhibitors6’7’10. Recent work specifically in Ph+ ALL examined developmental subtypes that align with earlier (Early-Pro) and later developmental (Late-Pro) B cell features, finding that the former was associated with poor overall survival upon treatment with the first-generation TKI imatinib8. Commitment to earlier or later stages of development has been associated with cooperating alterations in lineagedefining transcription factors (EBF1 deletion or deletions in IKZF1, PAX5, and CDKN2A, respectively), suggesting that developmental state adherence - and its associated therapeutic response - may be mutationally driven and static upon leukemic transformation8’9. However, other studies have nominated the potential for a leukemia's dominant developmental states to shift in response to therapeutic pressure. Illustratively, non-mutational mechanisms of chemotherapy resistance have been observed in ALL patient-derived xenografts (PDXs), whereby leukemia cells transiently adopt a dormant, stem-like state at MRD11; others have demonstrated post-treatment shifts in the abundance of dormant subpopulations mimicking earlier developmental stages7. It has also been suggested that TKI-resistant Ph+ ALL cells in a later developmental state proliferate by activating signaling that typically occurs downstreamof the pre-B cell receptor (pre-BCR), despite the absence of a functionally expressed pre-BCR in Ph+ ALL10’12. It remains unclear which attributes allow for persistence during remission and if mutational or developmental phenotypes are the dominant drivers of resistance.

[0233] Accordingly, resistance to ABL1 TKIs is multifactorial and extends beyond ABL1 resistance mutations, suggesting that informed strategies to convert deep remissions into cures may require incorporating orthogonal measurements of the non-genetic determinants of cellular state (e.g. via single-cell transcriptomics)13’14’15. However, there are limited studies describing how mutations participate (or clash) with these additional cellular features to drive persistence and clonal expansion under TKI pressure. Though recent evidence from our group and others indicates that some mutations are enriched in specific transcriptional backgrounds10’16’17’18’19, the relative importance of mutational and transcriptional drivers to MRD persistence and relapse is not known. Furthermore, there are significant technical challenges associated with isolating and profiling rare residual cells that have limited their characterization largely to mutational profiling - a problem affecting essentially all cancer types13’20’21. While MRD enumeration and mutational monitoring have been used to some clinical benefit22’23’24, the translational utility of understanding non-mutational attributes from these rare cells has yet to be demonstrated21. These constraints, coupled with the heterogeneity among MRD phenotypes both within and between patients, have historically made it difficult to nominate specific therapeutic strategies to combat MRD. Applicants and others previously proposed that direct interrogation of MRD cells to identify dependencies for individual patients could offer clinical benefit if approaches existed to define those dependences in "real-time"21’25. This would require a rapid strategy applicable to individual cells that could distinguish patients most likely to respond to one of several available therapeutic options.

[0234] Here, to better understand how both mutational and transcriptional variation coordinate to drive relapse within MRD, Applicants defined the biology of Ph+ ALL cells at different stages of treatment and across a diversity of models and human patients. Applicants reveal unique and targetable characteristics of Ph+ ALL MRD and nominate combination strategies to eradicate residual disease.ResultsModeling disease kinetics in response to combination TKI in Ph+ ALL PDX models

[0235] Although treatment with allosteric BCR::ABL1 inhibitors drives deep remissions in patients, nearly all will relapse if not consolidated with allogeneic stem cell transplantation.The recent development of asciminib (ABL001), an allosteric inhibitor ofBCR::ABLl, created the first opportunity to address whether dual inhibition of BCR::ABL1 could eradicate Ph+ leukemias (Figure 19A). Applicants combined orthosteric (ponatinib; 40 mg / kg / day) and allosteric (asciminib; 30 mg / kg / day) inhibitors in a diverse cohort of Ph+ ALL PDX models (n=13 models; 190 mice total) to assess how pre-existing clinical and molecular features would dictate response to sustained oncogene withdrawal within a statistically powered, phase Il-like preclinical trial (Figures 19A & 19B; see Methods **and Tables 1 & 2). All mice receiving ponatinib or combination therapy, and 92% of subjects receiving asciminib monotherapy who survived beyond one week achieved complete remission (CR), corroborating the dependence of these leukemias on BCR::ABL1 (Figures 25A & 25B). The durations of remission with ponatinib-based regimens exceeded those of asciminib monotherapy, but Applicants observed no difference between the combination and ponatinib monotherapy arms (p=0.70; Figures 19C & 25C). Notably, survival outcomes between mice on each treatment arm did not correlate with PDX line characteristics associated with inferior treatment response in other contexts, such as increased prior lines of therapy, IKZF1 deletion, and pre-existing ABL1 resistance mutations (Figure 25D). All mice were ultimately euthanized, either for disease progression or clinical toxicity. Even the 7 mice euthanized for clinical toxicity after achieving a durable response - three of whom maintained CR for >12 months on study (Figure 25B) - harbored residual ALL in the bone marrow and / or spleen when sacrificed. These data demonstrate that while single-agent ponatinib and combination therapy confer deep and prolonged clinical remissions, BCR::ABL1 inhibition alone was insufficient to fully eradicate human leukemias in vivo.Divergent mutational patterns upon oncogene inhibition in Ph+ B-ALL

[0236] To chart landscapes of genetic resistance to single agent and combination TKI in Ph+ ALL, Applicants sequenced 142 PDX samples (74 trial and 68 other TKI-treated leukemias) and examined patterns of acquisition within known driver mutations in ALL across multiple phases of treatment (Figure 26A; see Methods). In general, alterations in ABL1 or RAS pathway genes consistently emerged upon therapeutic pressure compared to mutations affecting B cell survival, lineage commitment, or cell cycle control (Figures 26A & 26B). Of relapsed leukemias, 35% harbored mutations in BCR::ABL1, frequently compound mutations involving T315I plus at least one other high-level resistance mutation (e.g., Y253H, F311L, F359V) or an activating mutation in STAT5A (collectively termed 'ABL pathway' mutations).A separate 24% relapsed with activating mutations in RAS pathway genes - specifically KRAS, NRAS, BRAF, and / or PTPN11 - representing emergent alternate pathway utilization in these oncogene-addicted leukemias (Figures 19D & 26C). Acquisition of driver pathway mutations was influenced by treatment arm - mice treated with asciminib predominantly acquired ABL pathway mutations at relapse, mice treated with ponatinib predominantly acquired RAS pathway mutations, and mice on the dual-treatment arm acquired mutations on either ABL or RAS pathways (Figure 26B). Samples harboring RAS pathway mutations were mutually exclusive with those involving ABL pathway mutations within each PDX line at both pretreatment and progression time points (Figure 19E). The remaining tumors (41%) harbored no driver mutations in either ABL or RAS pathway genes, and the majority of these (74%) had no apparent genetic lesions explaining phenotypic resistance by whole exome sequencing (Figures 19D, 26A, & 26C). These data suggest three recurrent patterns for resistance whereby leukemias progress on therapy with either ABL pathway, RAS pathway, or no discernible gain- of-function mutations.Ph+ ALL leukemic cells are defined by hybrid developmental states

[0237] Given the lack of discernible mutation-driven resistance in a substantial fraction of our cohort (Figures 19D & 26A), Applicants hypothesized that resistance to single-agent or combination TKI in Ph+ ALL may be understood best by characterizing both mutational and transcriptional state heterogeneity. To this end, Applicants applied single-cell RNA- sequencing (scRNA-seq) to define transcriptional states in Ph+ ALL and identify leukemic phenotypes associated with progression. Using Seq-Well S3, Applicants generated a dataset of 42,667 single-cell transcriptomes from 52 samples spanning 11 PDX lines from our phase II- like pre-clinical trial and 5 patients on a clinical trial testing dasatinib (a second-generation orthosteric BCR:: ABL 1 inhibitor) plus asciminib for previously untreated Ph+ ALL (Figure 20A; NCT02081378; see Methods). Applicants then performed consensus non-negative matrix factorization (cNMF) over each leukemia in this dataset to identify intratumoral gene expression programs (GEPs; Methods). Hierarchical clustering of the 126 GEPs defined across individual leukemias revealed 7 shared patterns (meta-GEPs, or "mGEPs") of covarying gene programming that were present in at least 8 samples (Figures 2B, 2C & S3 A; Table S4). Two mGEPs were defined by genes associated with active stages of the cell cycle (e.g., CENPF, MKI67, MCM6, E2F2) and another mGEP specifically associated with MYC activity (e.g., HSP90AB1, NME1). The remaining four mGEPs associated with various stages of B celldevelopment, either containing Pro-B cell genes (e.g., DNTT, CSGALNACT1), genes associated with later stages of B cell development — i.e., Pre-BII (e.g., CD38, IRF4), and Immature B (e.g., CD79A, HLA-DPB1) - or progenitor-associated genes co-expressed with Immature B genes (e.g., CD44, CSF1R and HLA-DQA1, IRF8). These data suggest that aspects of normal B cell development are captured as major axes of intratumoral transcriptional variation in Ph+ ALL.

[0238] In several cases, genes defining multiple B lineage developmental stages were enriched in the same GEP and co-expressed within individual leukemia cells (Figures 27A & 27B). Applicants next sought to better understand these stage-specific "hybrid" expression patterns by utilizing a supervised machine learning approach to resolve the relationship between leukemia cells and nonmalignant B cell development. To enable this comparison, Applicants first generated a reference dataset of human hematopoiesis from the bone marrow aspirates of healthy donors (n=7), profiling both sorted and unsorted fractions to ensure the proper B cell developmental populations were captured (Figures 28A & 28B; see Methods). By performing iterative clustering, Applicants identified 13 cell types spanning the HSC progenitor, myeloid, erythroid, and lymphoid lineages (n = 13,643 cells; Figures 28C & 28D); each cell type population contained cells from at least 6 of 7 donors (Figures 28E & 28F). To enable leukemic cell reference mapping and comparison, Applicants trained a random-forest (RF) classifier on the cell type-labeled reference scRNA-seq dataset using 10-fold cross- validation (Figures 20D & 29A; see Methods). Applicants ensured this model was cueing on biologically-relevant expression patterns by using permutation tests to identify the top 200 features needed to accurately classify single-cell transcriptomes, as well as testing its accuracy on an external scRNA-seq dataset (Figures 29B & 29C). Applicants then assigned individual B-ALL cells to their most likely developmental state using our RF classifier (Figure 20D). Across all malignant cells, the RF model assigned highest classification probabilities for the Pro-B cell type, followed by Pre-BI, Pre-BII, HSC, and Immature B cell types (Figures 20E & 20F); 1% of leukemia cells that classified into non-B lineage cell types, such as T cells, were poor quality and removed from downstream analyses (Figure 29D).

[0239] Corroborating our observations with NMF, marker genes that were restricted to individual stages of B cell development in healthy cells were routinely co-expressed in leukemia cells (Figure 20G). For example, within leukemic cells classified as Pro-B, Applicants observed a dominant secondary RF classification probability for an earlier (HSC)or later (Pre-BI, Pre-BII, Immature B) stage of B cell development. Applicants therefore characterized the transcriptional heterogeneity in Ph+ ALL as a continuum of hybrid states according to their non-Pro-B RF classification probability (Figure 20H). This revealed transcriptionally hybrid populations with underlying ProB-like gene-expression, co-expressed with either progenitor-like genes (HSC-hyb) or genes implicated in later developmental phenotypes (PreB-hyb or ImmatureB-hyb) (Figures 30A-C). Genes correlated with these prediction probabilities reflected markers of earlier and later stages of B cell development (Figures 29E & 30A; Table 4), and largely agreed with our unbiased NMF results (Figures 27A & 27C). All three hybrid populations were characterized by predicted utilization of canonical transcription factors (TFs) active in the healthy reference cell subsets (e.g., CREB1, MYC in HSC-hyb; E2F2, F0XM1 in PreB-hyb; IRF4, F0X03, CIITA in ImmatureB-hyb), as well as aberrant TF activity (e.g., IRF1, STAT1 in ImmatureB-hyb; Figure 30D; see Methods). Thus, anomalous co-expression of stage-associated genes in both primary patient samples and PDX models defines a hybrid development-like continuum in Ph+ ALL and implicates promiscuous, but still coherent, developmental transcriptional states.Hybrid development states are associated with treatment response and restricted mutation acquisition

[0240] Applicants next asked whether shifts in this hybrid development-like continuum associated with resistance to combination TKI. Overall, progression samples were characterized by decreased hybrid population diversity, suggesting a restriction toward a single hybrid state (Figure 21A). Differential expression analysis across all PDX tumors revealed genes included in the ProB-like (e.g., SOCS2, DNTT) and HSC-hyb (e.g., CD34, ID2, CD99) signatures enriched at pre-treatment while genes implicated in the more mature PreB-hyb (e.g., TCL1A, VPREB3, IGLL1) and ImmatureB-hyb (e.g., MS4A1, CD74, HLA-DRB1) signatures were up-regulated at progression, implicating a shift into later developmental stages (Figure 2 IB). However, not all PDX models shifted toward more mature hybrid transcriptional states at progression. DFAB-25157, which progressed with mutations in ABL1 (Figure 26C), remained dominated by ProB-like and HSC-hyb states at both pretreatment and progression compared to other PDX lines (Figures 21C & 31 A). Leukemias that progressed with RAS pathway mutations either contained a majority of cells expressing PreB-hyb and ImmatureB- hyb signatures at both pre-treatment and progression (CBAB-30198, DFAB-54880), or increased proportions of malignant cells with high PreB- and ImmatureB-hyb gene expressionat progression (DF AB-62208; Figure 26C). Notably, the two PDX lines that progressed with neither ABL nor RAS pathway mutations (CBAB-75914, CBAB-12402; Figure 26C) demonstrated the strongest shifts toward more mature hybrid developmental bins.

[0241] This enrichment for more mature phenotypes at progression was a strong departure from patterns seen in Ph+ ALL treated with chemotherapy or imatinib, where progression on therapy was driven by less mature or stem-like cells. Applicants sought corroborating evidence for this observation in our PDX trial samples using standard immunophenotyping approaches (Figure 3 IB; see Methods). Mirroring the transcriptional data, most pre-treatment leukemias harbored multiple subpopulations across the B cell developmental trajectory and showed a similar restriction in developmental state diversity at progression (Figures 21D & 31C). These immunophenotyping data also corroborated the overall enrichment of more developmentally mature phenotypes at progression (Figure 31C), specifically the predominance of more mature CD34-negative developmental phenotypes in leukemias that progressed with RAS pathway mutations or no mutations (p<0.001 from Dirichlet regression for both mutation group comparisons to ABL pathway-mutated leukemias; Figures 21E, 21F & 3 ID).

[0242] Applicants next sought direct clinical evidence for the relevance of developmentally-hybrid programs in resistance to combination TKI. Applicants prospectively collected serial single-cell measurements from the bone marrow of 2 patients (n=5 individual samples, n=7,649 cells; Figure 21G; Table 5) enrolled on a phase 1 trial testing dasatinib in combination with asciminib and prednisone. Clinical activity was assessed by the reduction in bone marrow BCR::ABL1 mRNA transcript levels after three cycles of treatment (day 85; NCT02081378). Samples from patient BIAB-16768 maintained a predominant population of ProB-like malignant cells over the course of treatment and entered remission before 85 days of treatment (3-log reduction in bone marrow BCR::ABL1 detected by qRT-PCR). By contrast, samples from patient DF AB-71417 rapidly shifted toward later developmental hybrid states (PreB-hyb and ImmatureB-hyb) by day 28 on therapy and failed to respond by day 85 (1-log reduction in bone marrow BCR: :ABL1 detected by qRT-PCR; Figures 21G & 21H). Combined with our PDX analysis, these results provide preliminary evidence that more mature developmentally-hybrid expression programs can drive resistance to dual ABL1 inhibition.Longitudinal monitoring of cell state and mutational co-evolution

[0243] Collectively, our data nominate 3 potential routes of resistance to ABL1 inhibition in Ph+ ALL: 1) mutational reactivation of ABL signaling in progenitor-like states, 2)mutational activation of RAS signaling in later-stage hybrid states, or 3) transcriptional shifts toward later developmental hybrid states without accompanying mutational alterations. To directly explore whether these routes are recoverable at multiple timepoints during ABL1 inhibition, Applicants next examined genotype-phenotype co-evolution by profiling single cells from pre-treatment, MRD (21 days on therapy), and progression in our PDX models, selecting individual leukemias that represent each putative mechanism of resistance (Figure 22A; DF AB-25157, ABL1 reactivation; DF AB-62208, RAS activation; CBAB-12402, no mutations). At each stage of therapy, Applicants profiled leukemia cells using SMART-Seq2 (SS2)-based scRNA-seq to increase information capture from low cell numbers at remission and to facilitate matched single nucleotide variant (SNV) detection in the same single cells. For these longitudinal studies, Applicants treated mice with single agent ponatinib (see Methods) since it performed equivalently to combination TKI therapy (Figure 19C) and is directly relevant to treatment being used in patients.

[0244] First, Applicants ensured the robustness of our RF hematopoietic developmental classifier on full-length, SS2 transcriptomes from both healthy (n = 421; same donors as Figure 28) and leukemic cells (n = 3,641; **Figure 32A; **see Methods). Using our RF framework, Applicants independently derived the leukemic cellular states in our SS2 dataset (Figures 32B- D; Table 31), finding they highly correlated with our Seq-Well-derived hybrid phenotypes - specifically in early progenitor (Progenitor-like vs. HSC-hyb) and more mature (PreB-like vs. PreB-hyb and ImmatureB-hyb) leukemic cell states (Figure 32E). Given this coherence, hereafter Applicants refer to Progenitor-like and PreB-like SS2 programs as HSC-hyb and PreB-hyb respectively for simplicity. Applicants next detected mutated transcripts identified from bulk DNA sequencing within individual cells from our SS2 data (Figure 33A; see Methods). The number of detected mutant transcripts in SS2 libraries was limited by the average expression of the corresponding gene, with higher rates of detection for RAS pathway single-nucleotide variants (SNVs; GNB1, NRAS, KRAS, PTPN1P) compared to ABL pathway SNVs (ABL1, S7A 75A) (Figure 33B). For highly expressed target genes, however, the proportion of single cells harboring mutations corresponded with the variant allele frequency measured in bulk sequencing of the same tumor (Figure 33C3), highlighting that SS2 provides sufficient SNV detection to capture the kinetics of RAS pathway mutations in our dataset. Furthermore, single-cell profiling enabled highly sensitive detection of rare malignant cells harboring mutations with less than 3% VAF from bulk sequencing, allowing comparisons ofdominant and rare subclones (Figure 33C). Finally, Applicants identified copy number variations (CNVs) in the SS2 profiles using inferCNV (see Methods). In combination with transcriptional state information, these data provided a detailed, high-resolution picture of the co-evolution of mutational and transcriptional heterogeneity in B-ALL single cells over the course of ponatinib treatment (Figures 22B & 33D-F).Cell state dictates fitness and restricts growth of RAS-mutant cells in remission

[0245] Using this high-resolution dataset, Applicants first evaluated changes in hybrid developmental state frequency between pre-treatment and residual cells in each model during treatment with ponatinib (Figure 22C). CBAB-12402 was transcriptionally dynamic and demonstrated a significant shift towards a dominant PreB-hyb phenotype among MRD cells that was conserved at progression, mirroring patterns seen in the larger PDX trial for this model (Figures 21C & 33F). Transcriptional states in MRD and progression leukemia cells from DF AB-62208 displayed a minor shift forward to stronger PreB-hyb expression compared to pre-treatment. DF AB-25157 was variable along the progenitor to mature phenotype continuum at both pretreatment and progression, driven by dominant HSC-hyb gene expression. A subset of cells from this model co-expressed HSC-hyb and PreB-hyb states in MRD, albeit at much lower levels than PreB-hyb scores in the other two models (Figures 33E-G).

[0246] Point mutations in NRAS and KRAS from the same leukemia cells revealed surprising dynamics across PDX models and stages of therapy (Figure 22D). Applicants detected very low frequency RAS mutations in CBAB-12402 at pre-treatment that were not enriched at progression, in agreement with bulk DNA sequencing data that did not identify actionable driver mutations (Figure 26C), thus implicating a "state-shift" only mechanism enabling progression. DF AB-62208 also harbored low-frequency KRAS and NRAS point mutations at pretreatment; a single NRAS-mutant, cycling cell was observed in remission and both KRAS-and NRAS-mutant clones expanded at progression (mirroring bulk sequencing data; Figure 26C), suggesting the preexisting PreB-hyb transcriptional state was permissive for expansion of RAS-mutant clones. In DFAB-25157, Applicants observed a significant increase in the proportion of KRAS mutant malignant cells in MRD (3 of 6 mice at MRD harbored identifiable RAS-mutant cells) compared to pretreatment leukemic cells, a finding Applicants confirmed using bulk DNA sequencing from a separate sample (Mouse 4H0, KRAS AF 0.75). This was surprising given that this model does not progress on therapy with emergent RAS mutations (Figures 22D & 26C). Indeed, considering both single-cell CNV and SNV clones(Figures 33D & 33E), Applicants found no evidence of outright genetically-driven clonal selection in DFAB-25157 despite the enrichment of RAS-mutant cells in remission (Figure 22E). In this case, our data suggest that RAS-family mutations in cells with a discordant developmental cell state permit survival (or persistence) in the context of ABL inhibition but confer a fitness disadvantage that suppresses their expansion.

[0247] Applicants next interrogated the single-cell transcriptomes of remission DFAB- 25157 cells to define mechanisms for this apparent state-genotype incompatibility. KRAS- mutant leukemic cells from DFAB-25157 at MRD upregulated genes that positively regulate senescence (e.g., CCL2, TOBI) and negatively regulate cell cycle (e.g., CDKN2A) compared to KRAS-mutant leukemia cells from all other time points and PDX lines (Figure 22F). To evaluate how this signature evolves over the course of therapy, Applicants scored individual cells for these upregulated senescence-associated genes (Senescence-like score; Table S8). KRAS-mutant clones with similar senescence-like signatures were present at pretreatment in cells with co-incident HSC-hyb phenotypes, whereas PreB-hyb KRAS-mutant leukemia cells across other treatment stages and PDX lines had low senescence-like scores (Figure 33G). These data suggest the fitness of RAS mutant clones is influenced by the compatibility of transcriptional state and genotype: the expression of senescence-implicated genes is restricted to HSC-hyb cells harboring RAS mutations, whereas RAS-mutant PreB-hyb cells remain capable of entering the cell cycle (Figure 22G). Therefore, despite activation of a mitogenic oncogene that contributes to resistance to TKI in multiple contexts, developmental states restrict the expansion of these genotypes, including during deep remissions.

[0248] As MRD genotypes alone could not predict clonal expansion driving progression, Applicants sought to identify what phenotypes persist in MRD and actively contribute to progression. Applicants binned each cell from MRD and progression into four fitness phenotypes based on their expression of senescence-like and cell cycle scores (Figure 22H). To our surprise, progression contained a significant accumulation of putatively cell cycle- arrested cells with higher senescence-like scores compared to MRD (p<0.001, KS statistic). Notably, Applicants also observed CNV subcl onal fitness plasticity in DFAB-25157, whose cells at MRD were characterized by high senescence-like scores. A cycling population of RAS- wildtype cells from one subclone emerged at progression (Figures 221 & 33H; p<0.01, Fisher's exact test), associated with an increased abundance of that subclone at progression (Figure 22E). In contrast, RAS-mutant cells from DF AB-62208, characterized by later developmentalphenotypes, were highly proliferative at progression (Figure 33H). Collectively, these data suggest that diverse Ph+ ALL genetic subclones can persist to progression and even clones with senescence-like phenotypes at MRD may expand with enhanced fitness to seed progression. Given the possibility of plasticity and the restrictions imposed by cell states on certain genotypes, these data suggest it may be difficult to predict from genetics alone the subclones that will ultimately seed relapse.Direct targeting of transcriptional programs in residual disease deepens remission

[0249] In light of this complexity, Applicants hypothesized that directly targeting transcriptional programs that enable persistence at MRD could overcome the diversity of subclones identified at remission. Using differential expression and gene-gene correlation (see Methods), Applicants identified three expression programs in remission that persisted to progression - a Pre-BCR Signaling program, closely aligned with the PreB-hyb state (e.g., IGLL1, VPREB3 a Stress / Autophagy program (e.g., HSPA1A, UBC), and an inflammatory program (e.g., EGR1, JUN, TNF,' Figure 23A; Table 8). The inflammatory program was evenly expressed across all leukemic cells in remission, a phenotype seen in other hematological diseases (28504724, 35618837; Figure 34A). The remaining expression programs were variable across MRD cells stratifying those high for the Stress / Autophagy cell state and those expressing the Pre-BCR Signaling program (Figures 34A & 34B). Applicants considered these variable programs to test the hypothesis that targeting specific expression programs could deepen remissions. These two variable gene expression programs split along fitness subpopulations, with leukemic cells harboring high Pre-BCR Signaling scores also scoring high for cell cycle, and leukemic cells with high Stress / Autophagy program scores enriched for senescence-like expression (Figure 23B).

[0250] Applicants next evaluated whether these two gene expression programs could be therapeutically targeted. Applicants paired ponatinib with either the FDA-approved SYK inhibitor, fostamatinib, to inhibit pre-BCR signaling in leukemic cells scoring highly for the Pre-BCR program, or the FDA-approved p38a MAPK inhibitor losmapimod, to target leukemic cells scoring highly for the Stress / Autophagy program given the co-enrichment of p38a MAPK activation with the Stress / Autophagy program and previous work supporting crosstalk between p38 signaling and autophagy / leukemic stem cell-related phenotypes (Figure 34C). As a combination control, Applicants compared transcriptional-state-directed combination therapy to dual oncogene targeting using ponatinib and asciminib (Figure 23C).Applicants selected two PDX lines that were enriched for either variable MRD expression program: DFAB-25157, which scored highly for the Stress / Autophagy program, and DF AB- 62208, which scored highly for the Pre-BCR signaling program and sat along the poised / cell cycle spectrum (Figures 23D, 23F & 34D). DFAB-25157 mice treated with combination losmapimod plus ponatinib showed a significant reduction in residual disease burden compared to dual oncogene suppression, a striking comparison as DFAB-25157 tumors consistently progressed with acquired mutations in ABL1 (Figures 23E & 34C). Analogously, DF AB- 62208 mice responded to ponatinib plus fostamatinib and had significantly reduced residual disease compared to dual oncogene suppression (Figure 23G). These data suggest that residual leukemia cells can be effectively targeted according to the specific transcriptional state governing persistence in remission.A biophysical workflow for low-cost rapid coupling of genotype to developmental state in leukemia cells

[0251] Our data support the importance of both mutations and overall cell state in determining leukemic cell fitness and therapeutic susceptibility at MRD. While mutations can be monitored in clinical workflows from residual leukemic cells, single-cell transcriptomics is currently difficult to scale due to the overall cost and time required for sample collection and analysis. Applicants sought a metric that would integrate complex transcriptional information from low-input MRD samples to enable rapid determination of leukemic cell state, compatible with downstream mutational profiling. Immunophenotyping strategies of developmental cell states, especially given the very low cell numbers at MRD, is likely to be highly challenging. Alternatively, cell size characteristically decreases as healthy progenitor cells progress from HSCs to pro-B to pre-B cells, putatively providing a label-free attribute with which to phenotype ALL cells. Applicants have previously shown that measurements of buoyant mass, as measured by the suspended microchannel resonator (SMR), can reveal changes in cell state. Buoyant mass (referred to hereafter simply as mass) can be measured from live single cells with a resolution near 50 fg, which is highly precise given that the average buoyant mass of a hematopoietic cell is ~75 pg. Further, Applicants have shown that coupling mass measurements to scRNA-seq from the same cell enables the determination of expressiondependent changes in cellular mass. Thus, Applicants hypothesized that underlying biophysical development-like phenotypes may be conserved and sufficient to rapidly capture the developmental state of a leukemia cell.

[0252] Applicants first determined whether mass can distinguish B cell developmental states in healthy donors. By performing paired SMR-SS2 on cells flow-sorted from healthy donors into Progenitor (CFU-L; 155 cells), Pro-B (122 cells), and Immature B (105 cells) gates, Applicants found that each stage of B cell development was characterized by distinct mass distributions, with decreasing cell mass along the B cell developmental trajectory (Figures 24A, 24B, 35 A & 35B). Within each B cell developmental stage, healthy cells with higher mass also scored highly for S phase or G2 / M phase cell cycle, a pattern seen across studies using SMRs within a specific cell type (Figure 35 A). Applicants found a strong relationship between each gene's dependence on RF prediction scores and matched cellular mass (r = 0.88 from Pearson correlation), indicating that genes highly associated with cell mass are also most correlated to healthy B cell developmental states (Figure 35C). Consistently, in leukemic cells, genes defining the HSC-hyb signature were most positively correlated with leukemic cell mass, and genes defining the PreB-hyb signature were most negatively correlated with cell mass (r = 0.90) (Figure 24C). Applicants validated this observation across 17 additional PDX samples at the bulk level showing that the average leukemic cell mass reflects the average RF predicted state (r = 0.66) and tracks with the progression-emergent mutations for each PDX (Figure 24D). Taken together, these data support mass as a meaningful surrogate for development-associated transcriptional state in leukemia cells.

[0253] Finally, Applicants evaluated how single-cell mass could pair with genotyping to further define developmental state and mutation compatibility (Figure 6E). Applicants compared mass distributions between RAS-mutant PDX lines with higher HSC-hyb and high senescence-like gene expression (DFAB-25157) and PDX lines with higher PreB-hyb gene expression (DF AB-62208 and DF AB-54880). State-genotype discordant HSC-hyb DFAB- 25157 cells were enriched for senescent-like scores and significantly higher mass than the more developmentally-mature and non-senescent DF AB-62208 and DF AB-54880, mirroring mass differences between healthy progenitor and immature B cells (Figure 24F). Furthermore, Applicants found a significant difference between the mass distributions of DFAB-25157 MRD cells compared to DF AB-62208 cells at MRD (Figure 24G), implicating that mass measurements reflect developmentally-relevant and therapeutically actionable heterogeneity in MRD for these leukemias (Figures 23D-G). Consequently, mass measurements appear to be sufficiently sensitive to distinguish differences in developmental state for leukemic cells, and,when assessed simultaneously with genotypic data from the same sample, may predict therapeutic susceptibility for targeting states in MRD.Discussion

[0254] Oncogene-directed therapy provides clear benefits to certain patient populations, yet it is equally clear that targeting cancers solely based on their mutational heterogeneity has an upper limit45’46’47. Indeed, our phase Il-like preclinical trial results reveal that even combinations of highly potent TKIs aimed at the same oncogene do not cure Ph+ ALL. While much of the preclinical and clinical data in CML and ALL have identified pathway reactivation through alterations in ABL1 as a primary mechanism of escape1’2’3’4, our data suggest alternative pathway activation through RAS alterations also drives resistance in a significant fraction of cases. Mirroring patterns seen in patients5, our trial also shows that a large fraction of mice engrafted with patient-derived leukemias (up to 40%) progress without a clear genetic driver, warranting the exploration of alternate therapeutic strategies for these cases.

[0255] Transcriptional phenotypes have been described in AML16, CML13, and ALL8, and recent studies suggest that patients with more progenitor-like leukemia cells have a worse overall prognosis and tend to respond poorly to therapy. In ALL specifically, a recent study showed that leukemias enriched for progenitor-like states have worse outcomes on imatinib8. Our data suggest that lineage plasticity is relatively common in response to 3rd generation and combination TKI therapy, with resistant leukemia cells most frequently mimicking later stages of B cell development. This contrasts with most settings where, even in solid cancers, a canonical response to therapy is the enrichment of less differentiated cell states48’49. Moreover, Applicants demonstrate the importance of defining cell state and mutational associations - despite myriad mutational routes that might be predicted to confer resistance, our data suggest that specific transcriptional backgrounds may restrict leukemias to distinct subsets of escape mutations. Though these associations will need to be learned in larger cohorts and for each specific disease, this framework may represent a strategy for prioritizing the permissible transcriptional state / mutational convergences within oligo / polyclonal populations that can drive progression.

[0256] While there is agreement on the clinical and therapeutic importance of understanding MRD, the phenotypes of the residual cells responsible for seeding progression and how to best target them remains an outstanding question in the field owing to several technical challenges21’25. In this regard, Ph+ ALL is a tractable system, as it is feasible to isolateMRD from either blood or bone marrow of patients or xenografted mice in adequate numbers to allow for single-cell transcriptomics in addition to DNA sequencing. Applicants found that matched genotypic and phenotypic profiling of rare MRD cells was critical for identifying three key insights about the biology of MRD and the translational potential of targeting it prior to relapse. First, the conventional wisdom proposes that not all cells at MRD can seed relapse, especially those that have exited the cell cycle or are otherwise classified as "unfit"21’25. In contrast, Applicants find that some CNV-defined clones expressing senescence-like genes at MRD can re-enter the cell cycle and contribute to progression. Of note, a similar phenotype has also been observed in AML treated with chemotherapy50. Second, our discovery that senescent clones harboring RAS mutations were enriched in residual disease but did not contribute to relapse highlights the importance of understanding the cell state of mutant cells. This observation complicates current MRD evaluation strategies, as information about genotype alone will likely be insufficient to predict relapse for specific leukemias. Third, Applicants show that co-targeting tumor-specific transcriptional programs in remission outperforms additional targeting of the same oncogene, at least with current therapeutics. This finding provides a translational rationale for identifying transcriptional phenotypes in residual disease to inform the rational selection of combination strategies. The importance of targeting cell state likely extends to other cancers where a central oncogene can be deeply inhibited, resulting in relapses that have acquired an alternate histology, including small cell relapse after androgen receptor inhibition in prostate cancer51, squamous cell and small cell transitions after EGFR inhibition in lung adenocarcinoma52’53, and estrogen receptor positive relapse after HER2 blockade54

[0257] Applicants note that the influence of an intact immune system on the developmental dynamics of Ph+ ALL is not well defined and represents a liability of our approach interrogating PDX models of leukemia in NSG hosts. Applicants mitigated this by confirming our PDX results in serial measurements from patient bone marrow, but future efforts should include the use of humanized xenograft models and additional evaluation of primary patient specimens. Nevertheless, our identification of a central role for developmental state in Ph+ ALL has had immediate clinical implications. Our phase 1 clinical trial of dual oncogene targeting (NCT03595917) completed accrual55and reopened as a phase 2 trial incorporating early introduction of the CD3xCD10 bispecific antibody blinatumomab (anti-CD3xCD19 bispecific antibody), which should maintain activity across the developmental statesApplicants have defined in MRD and relapse. Importantly, blinatumomab has demonstrated promising clinical activity in clearing residual disease in patients intended for consolidative allogeneic hematopoietic stem cell transplantation56’57.

[0258] Evaluating complex, non-mutational biomarkers may have significant clinical challenges. scRNA-seq is not yet a clinically-scalable assay, nor is it readily interpretable on a short time-scale. For translation to clinical workflows, it will be critical to develop diagnostics that are able to assess a sample's genotype and relevant phenotype with reasonable throughput and interpretability. For remission profiling specifically, this is further complicated by the requirement for use with low-input samples. Owing to the low-input and non-destructive nature of the SS2-SMR measurement41, Applicants were able to acquire a unique dataset that directly links cellular mass to leukemic developmental state. These data establish that assessing complex, non-mutational biomarkers may be possible using mass as a relatively simple integrative cellular property. Our matched SMR / scRNA-seq data from normal bone marrow hints that mass variation may extend to other hematopoietic lineages as well so this approach may be applicable in diseases with significant developmental heterogeneity such as AML16. Applicants speculate that additional features of clinical utility in different disease contexts could come from other integrative single-cell properties such as morphology58.

[0259] In sum, Applicants find transcriptional state controls the fitness of individual clones in MRD and dictates the landscape of progression on TKI in Ph+ ALL. Applicants highlight the need to understand and monitor both mutational and transcriptional features in clinical pipelines to properly evaluate individual clones for their potential to drive relapse. Applicants functionally establish the paramount importance of cell state in this context and suggest it should be prioritized for targeting in conjunction with driver oncogenes. In agreement with recent studies in solid cancers59’60’61, our work in leukemia makes it apparent that therapies intended to convert remissions to cures should consider monitoring and targeting features outside of traditional mutational biomarkers62.ResultsData Availability

[0260] The scRNA-seq data and SMR data reported in this paper will be deposited in a central data sharing repository (Genomic Data Commons) under the NCBI Database of Genotypes and Phenotypes (dbGaP). scRNA-seq digital gene expression matrices, metadata, and interactive visualization tools will additionally be available through the Alexandria Project,a Bill & Melinda Gates Foundation-funded portal (part of the Single Cell Portal hosted by the Broad Institute of MIT and Harvard). Code used for analysis will be available upon request. Generation and Use of PDXs

[0261] Primary bone marrow and peripheral blood specimens were collected from patients with leukemia at the Dana-Farber Cancer Institute, Brigham and Women's Hospital, and Boston Children's Hospital for xenotransplantation. Additional PDXs that had already been established through the Public Repository of Xenografts (PRoXe) were utilized. De-identified patient samples were obtained with informed consent and xenografted under Dana- Farber / Harvard Cancer Center Institutional Review Board (IRB)-approved protocols. Nod.Cg- PrkdcscidIL2rgtm IWjU zS (NSG) mice were purchased from Jackson Laboratories and handled according to Dana-Farber Cancer Institute Institutional Animal Care and Use Committee-approved protocols. Salient PDX line metadata are provided in Tables SI & S2. In vivo therapeutic studies

[0262] Viably frozen Ph+ ALL xenograft cells were thawed and changed into IX PBS before tail-vein injection at 0.5-2.0*10A6Acells per mouse. Engraftment was monitored by weekly peripheral blood flow cytometry beginning three weeks after injection. Blood was processed with Red Blood Cell Lysis Buffer (Qiagen #158904; Hilden, Germany) and stained with antibodies against human CD45 (APC-conjugated, eBioscience #17-0459-42; San Diego, CA, USA) and human CD19 (PE-conjugated, eBioscience #12-0193-82) in IX PBS with EDTA (2mM). Flow cytometry data were analyzed using FlowJo software (BD Biosciences; Ashland, OR, USA). Upon engraftment - when at least 10% of cells were positive for CD45 and CD19 - mice within each PDX line underwent 1 :2:2:4: 1 randomization to the following arms and initiated treatment within two days: (1) sacrifice for baseline tissue interrogation; (2) ponatinib (Selleckchem #S1490; Houston, TX, USA; constituted in 25mM citrate buffer, pH 2.75) 40mg / kg via oral gavage (OG) daily; (3) asciminib (NVP-ABL001, Novartis Pharmaceuticals; Basel, Switzerland; constituted in HC1 0.1M, PEG300 30%, Solutrol HS15 6%, NaOH 0. IM, sodium acetate buffer pH 4.7 lOmM) 30mg / kg OG twice daily; (4) ponatinib 40mg / kg OG twice daily plus asciminib 30mg / kg OG BID; and (5) vehicle (alternating doses of vehicle used for ponatinib and asciminib, at equivalent volumes). One mouse per active treatment arm per PDX line was sacrificed on day 7 of treatment for pharmacodynamic assessment. The remaining mice continued daily treatment under monitoring with biweekly peripheral blood flow cytometry until progression (defined as peripheral blood involvement ofat least 10% on two consecutive assessments at least one week apart), weight loss of greater than 20% from pre-treatment baseline, or clinical manifestations of advanced disease, including but not limited to ruffled fur, hunched posture, hind limb paralysis, or lethargy. Progression or toxicity as defined above triggered humane euthanasia by CO~2~ asphyxiation, necropsy to ascertain cause of death, and post-mortem harvest of peripheral blood, bone marrow, and any soft tissue masses. Additional in vivo studies involved treatment with nilotinib (Selleckchem #S1033), which was constituted in N-methyl-2-pyrrolidone (10%) in polyethylene glycol (PEG)-300 (90%) and dosed at 50mg / kg OG twice daily.

[0263] Studies to define the in vivo activity of combination therapies targeting the biology of MRD within individual PDX lines DF AB-62208 and DFAB-25157 utilized the same xenotransplantation and engraftment monitoring scheme as previously described and the following drugs: ponatinib (as above), asciminib (as above), fostamatinib (Selleckchem #S2206-50mg), constituted in 0.1% carboxymethylcellulose sodium, 0.1% methylparaben, and 0.02% propylparaben (pH 6.5) and dosed at 25mg / kg OG thrice daily, and losmapimod (Selleckchem #S7215-50mg), constituted in 1% DMSO in methylcellulose and dosed at 20mg / kg via the intraperitoneal (IP) route daily. Upon engraftment (>10% leukemia involvement of peripheral blood), individual mice underwent live femoral bone marrow aspirates under anesthesia with inhaled isoflurane delivered via precision vaporizer and underwent 1 : 1 : 1 randomization to the combination of ponatinib and asciminib, ponatinib and fostamatinib, or ponatinib and losmapimod. Animals initiated treatment within 48 hours of engraftment and continued treatment for 21 days ± 3 days, at which point they underwent humane euthanasia, necropsy, and immediate post-mortem recovery of peripheral blood and bone marrow from the femur contralateral to that which was aspirated upon engraftment.Human donors for reference

[0264] Normal human bone marrow aspirates were obtained from donors who provided informed consent for tissue banking and research under Dana-Farber / Harvard Cancer Center IRB protocols and were undergoing bone marrow harvest for unrelated hematopoietic stem cell transplantation recipients. Briefly, bone marrow was collected into a Baxter bone marrow harvest collection system with diluent consisting of sodium heparin in lactated Ringers solution. Bone marrow was heparinized at a final concentration of 15-20 units / mL and filtered inline using 200pm and 500pm filters. Bone marrow mononuclear cells from the heparinized, filtered product were isolated via density gradient centrifugation (Ficoll-Paque, ThermoFisherScientific #45-001-749) and subsequently underwent fluorescence-activated cell sorting (FACS) to isolate hematopoietic developmental subpopulations for Seq-Well S3 and SS2 single-cell transcriptomic profiling (see Methods Details).Phase I clinical trial

[0265] Serial primary blood and bone marrow specimens were obtained from appropriately consented patients treated on a phase I, investigator-initiated clinical trial (NCT03595917) of asciminib (ABL001) in combination with dasatinib plus prednisone for adults with newly diagnosed Ph+ ALL or chronic myelogenous leukemia in lymphoid blast phase (CML-LBP). Some patients cross-consented to a Dana-Farber Cancer Institute tissue banking protocol permitting additional evaluation of primary specimens. Bone marrow was obtained at screening and after each 21 -day cycle through the first four cycles. Peripheral blood was obtained at screening and on days 2, 4, 8, 11, 15, and 22 (±2 days) of cycle 1. Both bone marrow and peripheral blood were collected into EDTA vacutainer tubes prior to mononuclear cell isolation per standard protocols. Bone marrow and peripheral blood underwent clinical quantitative real time PCR for BCR::ABL1 mRNA according to the BCR::ABL1 isoform detected at screening (pl 90 or p210). Curated sets of Ph+ ALL clinically annotated specimens underwent evaluation by scRNA-seq (Seq-Well S3; salient donor metadata provided in Table S6).Quantifying BCR::ABL1 mRNA in PDX peripheral blood with qRT-PCRBCR: :ABL1 mRNA levels were measured via quantitative real-time PCR (qRT-PCR) of serial peripheral blood specimens from PDX models to track kinetics of response and progression. Briefly, xenografted mice were phlebotomized for lOOpL by submandibular vein laceration every two weeks. Blood was stored in RNAProtect tubes (Qiagen #76544). mRNA was isolated using the RNeasy Protect Animal Blood Kit (Qiagen #73224) and quantified using the iScript One-Step RT-PCR Kit with SYBR Green (Bio-Rad #170-8893) on a Bio-Rad CFX96 Thermal Cycler. Synthesis of cDNAs was performed with random hexamers. Amplification of cDNAs was performed using iTaq Universal SYBR Green Supermix (Bio-Rad #172-5125) and the following oligomers:BCR-ABL isoform pl90 forward: CAACAGTCCTTCGACAGCAG (SEQ ID NO: 1) BCR-ABL isoform pl90 reverse: CCCTGAGGCTCAAAGTCAGA (SEQ ID NO: 2) BCR-ABL isoform p210 forward: TCCGCTGACCATCAATAAGGA (SEQ ID NO: 3) BCR-ABL isoform p210 reverse: CACTCAGACCCTGAGGCTCAA (SEQ ID NO: 4)Positive control reagents for each isoform were pl 90 clonal control RNA (Invivoscribe #4- 089-2800) and mRNA isolated from the BCR::ABL1 p210-positive cell line K562.Quantifying BCR::ABL1 mRNA in primary patient peripheral blood with qRT-PCR

[0266] BCR::ABL1 mRNA was quantified in the peripheral blood of patients treated on clinical trial NCT03595917 via CAP / CLIA-approved clinical BCR::ABL1 qRT-PCR performed in the clinical molecular laboratory of Brigham and Women's Hospital (Boston, MA).Targeted DNA Sequencing

[0267] PDX models underwent mutational profiling with targeted panels. Leukemia cells were enriched from fresh primary PDX bone marrow or peripheral blood via immunomagnetic enrichment for human B cells using human CD 19 MicroBeads (Miltenyi Biotec #130-050-301; Gaithersburg, MD, USA). DNA was extracted using the DNeasy Blood & Tissue kit (QIAGEN #69504) and fluorometrically quantitated using the Qubit dsDNA HS assay kit (Invitrogen #Q32854; Waltham, MA, USA) prior to use in next-generation sequencing library preparation.

[0268] A hybrid-capture target enrichment panel targeting the full coding sequences of 183 genes selected based on the presence of recurrent mutations in hematologic malignancies was utilized to profile most PDX models at baseline, on-treatment, and at end of study (as previously described).A63AAn amplicon-based clinical sequencing panel targeting hotspot regions of the oncogenes and most of the coding regions of tumor suppressor genes recurrently implicated in hematologic malignancies (total 93 genes) was employed for a subset of PDX models.A64AA custom amplicon-based deep sequencing panel targeting 23 genes implicated in in B-ALL treatment resistance (ArcherDX; Boulder, CO, USA) was employed to profile PDXs progressing after BCR: :ABL1 inhibition.Whole Exome Sequencing (WES) sample preparation

[0269] PDXs that progressed in absence of treatment-emergent driver alterations detected by targeted sequencing underwent whole exome sequencing using the SureSelect Human All Exon v5 kit (Agilent Life Sciences; Santa Clara, CA, USA). Briefly, lOOng of genomic DNA from each leukemia specimen as well as a control cell line (CEPH 1408) and a tail clipping from a non-xenografted NSG mouse were fragmented to 250bp on a Covaris Ultrasonicator (Woburn, MA, USA). Size-selected DNA fragments were ligated to xGen vl UDI-UMI9 adaptors (Integrated DNA Technologies; Coralville, IA, USA) during automated library preparation with a Biomek FXp liquid handling robot (Beckman Coulter; Indianapolis, IN,USA). Libraries (250ng per sample) were pooled to 750ng and captured with the SureSelect Human All Exon v5 bait set. Captures were pooled and sequenced on a HiSeq 3000 (Illumina; San Diego, CA, USA).Flow sorting of from healthy human bone marrow aspirates and PDX tumors

[0270] Approximately 106cells per sample were resuspended in PBS with 4,6-diamidino- 2-phenylindole (DAPI; 0.75pg / mL) as a dead cell marker. For cell surface staining, PBS- washed cells were blocked with Fc blocker for 10 min on ice and then stained with the antibodies listed in Table S10 at the manufacturers' recommended concentrations or with an isotype control for 25 min on ice. Cells were then washed and resuspended in chilled PBS containing 0.75pg / mL of DAPI to exclude dead cells. For annexin V staining, annexin V binding buffer (BD Biosciences) was used instead of PBS, and 7-aminoactinomysin D (7- AAD; BD Biosciences) instead of DAPI. Phycoerythrin (PE)-labelled annexin V was purchased from BD Biosciences. Acquisition was performed on a LSR Fortessa flow cytometer (BD Biosciences). Fluorescence-based cell sorting was performed on a FACSAria II (BD Biosciences). FACS data were analyzed with FlowJo software (FlowJo).

[0271] Cells expressing B cell lineage-defining surface proteins were enriched by FACS on a BD FACSAria II cell sorter (BD Biosciences; Franklin Lakes, New Jersey, USA) based on staining with antibodies targeting the following markers: Annexin V, CD45, CD34, CD 10, CD 19, CD20, and CD22. Healthy and immunophenotyped subpopulations were defined as in Figures S4A & S7B. Lymphoid progenitor sub-populations then underwent scRNA-seq via Seq-Well S3 and SS2.Sample preparation for scRNA-seq of clinical and PDX samples

[0272] Applicants used the Seq-Well S3 platform for massively parallel scRNA-seq to capture transcriptomes of single cells on barcoded mRNA capture beads.A31ABriefly, a singlecell suspension of 15,000 cells in 200pL RPMI media supplemented with 10% FBS was loaded onto single arrays containing barcoded mRNA capture beads (ChemGenes). The arrays were sealed with a polycarbonate membrane (pore size of 0.01pm), before undergoing cell lysis and transcript hybridization. The barcoded mRNA capture beads were then recovered and pooled for all subsequent steps. Reverse transcription was performed using Maxima H Minus Reverse Transcriptase (Thermo Fisher Scientific EP0753). Exonuclease I treatment (NEB M0293 L) was used to remove excess primers, followed by Second Strand Synthesis using a primer of eight random bases to create complementary cDNA strands with SMART handles for PCRamplification. Whole transcriptome amplification was carried out using KAPA HiFi PCR Mastermix (Kapa Biosystems KK2602) with 2000 beads per 50-pl reaction volume. Libraries were then pooled in sets of eight (totaling 16,000 beads), purified using Agencourt AMPure XP beads (Beckman Coulter, A63881) by a 0.6* solid phase reversible immobilization (SPRI) followed by a l x SPRI, and quantified using Qubit hsDNA Assay (Thermo Fisher Scientific Q32854). The quality of whole transcriptome amplification (WTA) product was assessed using the Agilent High Sensitivity D5000 Screen Tape System (Agilent Genomics) with an expected peak at 800 base pairs tailing off to beyond 3000 base pairs and a small / nonexistent primer peak.

[0273] Libraries were constructed using the Nextera XT DNA tagmentation method (Illumina FC-131-1096) on a total of 750pg of pooled cDNA library from 16,000 recovered beads using index primers with format as previously described.A31ATagmented and amplified sequences were purified at a 0.6x SPRI ratio yielding library sizes with an average distribution of 300 to 750bp in length as determined using the Agilent High Sensitivity D5000 Screen Tape System (Agilent Genomics). Two arrays were sequenced per sequencing run with an Illumina 75 Cycle NextSeq 500 / 550 v2 kit (Illumina FC-404-2005) at a final concentration of 2.4pM. The read structure was paired end with Read 1 starting from a custom Read 1 primer containing 20 bases with a 12-bp cell barcode and 8-bp unique molecular identifier (UMI) and Read 2 containing 50 bases of transcript sequence.Sample preparation for paired SMR mass profiling and SMART-Seq2

[0274] For all PDX and healthy bone marrow samples, cells were adjusted to a final concentration of 2.5* 105cells / ml to load single cells into the mass sensor array and record single-cell mass measurements, as previously described41’65. In order to exchange buffer and flush individual cells from the system, the release side of the device was constantly flushed with PBS at a rate of 15pL per minute. Upon detection of a single-cell at the final cantilever of the SMR, as indicated by a supra-threshold shift in resonant frequency, a set of three- dimensional motorized stages (ThorLabs) was triggered to move a custom PCR-tube strip mount from a waste collection position to a sample collection position to retrieve the cell. Each cell was dispensed in approximately 5pl of PBS into a PCR tube containing 5pl of 2x TCL lysis buffer (Qiagen) with 2% v / v 2-mercaptoethanol (Sigma) for a total final reaction volume of 1 Opl. After each 8-tube PCR strip was filled with cells, the strip was spun down at 1,000 gfor 30 seconds and immediately snap-frozen on dry ice. Following collection, samples were stored at -80°C prior to library preparation and sequencing.

[0275] Single-cell lysates were compiled from independent collections upon thawing and transferred into wells of a 0.2mL skirted 96-well PCR plate (Thermo Fisher Scientific). scRNA-seq libraries were generated using SMART-Seq2 protocol66. Briefly, cDNA was reversed transcribed from single cells using Maxima RT (Thermo Fisher Scientific) and whole transcriptome amplification (WTA) was performed. WTA products were purified using the Agencourt AMPure XP beads (Beckman Coulter) and used to prepare paired-end libraries with Nextera XT (Illumina). Single cells were pooled and sequenced on a NextSeq 550 sequencer (Illumina) using a 75 cycle High Output Kit (v2.5) with a 30bp paired end read structure.PDX in vivo studies: survival analysis on treatment arms and with pretreatment clinical risk stratification metadata

[0276] Analyses fitting a Cox proportional hazards model for overall survival (OS) and progression-free survival (PFS) outcomes on treatment arms and pretreatment clinical risk stratification categories were performed using the survival package in R.A67AThe following pre-clinical features included: IZKF1 deletion, 9p deletion, hyperdiploid karyotype, gain of chromosome 21, presenting white blood cell count, age, sex (if age <18 years), race, phase of disease, number of prior therapies, and pre-existing ABL1 mutation(s). Hazard ratios and p- values for PFS within pretreatment clinical risk categories were generated relative to the lowest risk group in each category (Figure 25D).WES alignment and variant calling

[0277] Pooled sequenced WES samples were demultiplexed using Picard tools. Read pairs were aligned to the hgl9 reference build using the Burrows-Wheeler AlignerA68AData were sorted and duplicate-marked using Picard tools. Alignments were refined using the Genome Analysis Toolkit (GATK)A69,70Afor localized realignment around small insertion and deletion (indel) sites. Mutation analysis for single nucleotide variants was performed with MuTect vl.l.4A71Aand annotated by Variant Effect PredictorA72AIndels were called using the SomaticIndelDetector tool of the GATK. Copy number variants (CNVs) were identified using RobustCNV for autosomes.A73ADetected alterations are reported in Figure 26A. scRNA-seq sequencing alignment and quality control

[0278] Sequenced Seq-Well BCL files were demultiplexed into individual sample FASTQs for Read 1 and Read 2 using the bcl2fastq pipeline on Terra, as previously described.The resultant paired read FASTQs were aligned to the hgl9 genome using the cumulus / dropseq tools pipeline on Terra maintained by the Broad Institute using standard settings, generating a genes by cells count matrix for each sample.A74ALow quality cells were filtered using nGene<200, nUMI<500, and percent mitochondrial transcripts<30% thresholds before merging samples; genes were filtered if they were not expressed in at least 10 cells.

[0279] Sequenced SS2 BCL files were similarly demultiplexed using bcl2fastq and aligned to the hgl9 genome using publicly available scripts on Terra (github.com / broadinstitute / TAG- public). Total gene counts and transcript per million (TPM) matrices were filtered to remove low quality cells with <15% transcriptome mapping, 2,000 genes, and 45,000 mapped reads, before continuing analysis. Genes expressed in fewer than 10 cells, as well as long non-coding RNAs and unique hgl9 reference-build variants were removed before downstream analysis. Human healthy bone marrow reference cell type clustering and visualization

[0280] After QC filtering, 13,643 high quality cells from 7 healthy human bone marrow donors were analyzed in Seurat v2.3.4 to classify hematopoietic cell types.A75AAfter normalization, the top 1,500 highly dispersed variable genes were selected using the meanvariance plot method in Seurat's FindVariableFeatures function. ScRNA-seq data was scaled over highly variable genes and used as input for PCA analysis. The top significant PCs, as defined by the JackStraw test (top 25 PCs), were used as input for building a SNN graph to cluster cells by their (k=35) nearest neighbors and for t-SNE visualization of clusters. Given the shared, continuous hierarchy of covarying gene expression in hematopoietic development, broad cell types (progenitor, myeloid, erythroid, B cell lineage, pDCs, T cells, and Plasmablasts) were called based on their differentially expressed genes (identified using the Wilcox test in Seurat's Find AllMarkers function), and subset into individual Seurat objects for a second round of clustering to resolve the final 13 cell types defined in Figure 28. Cell type annotations were post-hoc validated based on biased or exclusive expression of known marker genes (Figure 28D).

[0281] SS2 healthy reference cell types were called by their confident random forest prediction probabilities (see next section) and examination of marker genes to provide further support of cell type identification (Figure 32B). Cell type clusters were visualized using SPRING, a tool that generates force-directed layouts from kNN graphs to visually preserve hierarchical relationships between cell types76.Unbiased identification of consensus intratumoral gene expression programs with NMF

[0282] Applicants sought to identify common axes of covarying intratumoral gene expression within all Ph+ ALL tumors in our dataset. First, Applicants ran consensus NMF (cNMF) on each tumor in our dataset (n=52 total samples, defining bone marrow and spleen samples from the same mouse as individual tumors)77. For this analysis, Applicants selected a consensus 1,489 variable genes across all tumors by first identifying the top 2,500 variable genes within each individual tumor using the variance standardized transformation method in Seurat v5.0.2 FindVariableFeatures function. To ensure consensus variable gene selection was not biased by PDX line- or patient-specific variable genes, as some models or donors had more tumors sampled than others, Applicants initially selected the top 2,000 median weighted variable genes across tumors within a PDX line or patient, and then chose the top 2,000 median weighted variable genes across all PDX line and patient median gene lists. 511 of these top 2,000 variable genes were removed based on non-zero expression across all 52 tumors.

[0283] cNMF (1,000 iterations) was performed on the counts matrices of each tumor utilizing the consensus variable gene list over a range of k=3-9. All stable solutions of k, defined by a cNMF solution silhouette score>0.8 across iterations, were evaluated for optimal k selection using the following heuristics. Applicants first hierarchically clustered the Jaccard Similarity of the top 50 genes from each factor across all stable k solutions; under-clustered k solutions were nominated based on factors that contained genes split across clusters that were hierarchically clustered in higher k factorizations, and over-clustered k solutions were nominated based on the presence of factors that did not hierarchically cluster with lower k factorizations or split genes across multiple lower k factors.A78ATo further evaluate these hypothesized over- or under-clustered k solutions, Applicants scaled the data and ran UMAP projections over the top 50 genes from each factor for each stable k solutions. Applicants used Seurat's AddModule Score function over the top 50 genes from each factor to assess whether under-clustered factors convolved expression across UMAP subclusters of optimal k solutions, or whether over-clustered factors scored highest in the same subcluster of cells or mostly strongly defined 1-2 cells ("junk" factor). Finally, Applicants assessed significant Pearson correlation of the top 50 genes in each optimal k factor over an expression-binned bootstrapped null distribution as previously described,A79Aremoving factors that were not significantly correlated (typically "activity "-like continuous programs in UMAP projections that, upon inspection, actually contained sparsely expressed genes of redundant biological annotations to other factors within that k solution). Factors from the selected optimal k that containedsignificantly correlated genes were labeled as "intratumoral gene expression programs" or GEPs, and collated for downstream intertumoral comparisons across the entire tumor cohort. Examples of intratumoral GEPs from representative PDX and patient tumors are shown in Figure 27B.

[0284] From performing intratumoral cNMF on 52 tumors, Applicants identified 166 intratumoral GEPs. Applicants excluded outlier GEPs by constructing a kNN graph (k= 15) and filtered 40 intratumoral GEPs using an elbow-based filtering criterion over kNN distances of each individual GEP to its nearest neighbor. The remaining 126 intratumoral GEPs were hierarchically clustered using Ward.D clustering over their cosine similarity to reveal 7 meta- GEPs or "mGEPs", which Applicants interpret as shared intratumoral gene covariation across at least 8 individual tumors (Figure 27A). To interpret shared gene covariation across each identified mGEP, Applicants isolated the top 30 median gene loadings across intratumoral GEPs within a given mGEP cluster (Figures 3C & S3 A; Table S4).Training and interpreting the random forest classifier

[0285] Random forest is an ensemble machine learning method used for both classification and regression. Like other ensemble models, random forests combine multiple weak classifiers, in this case shallow decision trees, to make predictions. In this work, a random forest was used for classification. Here, Applicants interrogate aberrant developmental hierarchies in ALL by using random forests to predict the nearest cell type from the normal B-cell lineage for single cells from Ph+ ALL samples. There are inherent advantages to random forests for the Ph+ ALL classification task. Importantly, ensemble classifiers, like a random forest, provide a distribution of class probabilities reflecting the similarity of each cell to each cell type the model was trained on. This is done by calculating the proportion of trees voting for a cell type for each given observation. To generate a single prediction for a cell, the highest-class probability becomes the prediction. The higher the probability of the chosen class, the more transcriptionally similar the cell is to that stage of B cell development. The distribution of class probabilities itself can be used to understand the certainty - or uncertainty - of a prediction. Applicants leveraged this measure of uncertainty in predictions to evaluate how well a tumor cell fits a specific stage in B cell lineage (Figure 2H). A tumor cell with a more uniform distribution of probabilities over classes likely shares transcriptional features with many a wider range of stages of B cell development, potentially indicating a more aberrant cell from normal development. Second, ensemble approaches tend to be more robust to overfitting,which is necessary when applying a model trained on sorted, healthy populations of cells to evaluate aberrant leukemic cells. Finally, because random forests are nonparametric models, they also are highly flexible to input feature scale and variance. This makes the approach particularly suited to raw count matrices output by various scRNA-seq technologies used.

[0286] Here, Applicants trained a random forest on sorted cells from the B cell lineage using 15,000 genes with detected expression in more than 10 cells as input features. Random forests were implemented using R version 3.5.1 using the caret package for training infrastructure.A80AThe ranger implementation of random forests was used.A81AHyperparameter search over ranger parameters (the number of randomly selected features considered for splitting at each tree node and the rule used for splitting) was done via 10-fold cross-validation (CV). The model achieved an accuracy of 94±0.006% on 10-fold CV with optimal parameters. The final model used the full training set of 13,643 cells. Results of 10- fold cross validation are provided in Figure 29A. The model was also evaluated on an external testing set of Seq-Well generated healthy bone marrow scRNA-seq transcriptomes16, and achieved performance of average AUC=0.99 over all 13 cell types (Figure 29C). To interpret features being used to make predictions by the classifier, Applicants used permutation importance tests. Permutation importance measures the impact of randomly shuffling feature values on the performance of a model measured as accuracy and decrease in Gini impurity. Specifically, a computationally accelerated heuristic method was used that constructs a null distribution from features that have importance values close to zero, limiting the need for randomly shuffling all features independently to evaluate significance.A82AThe results of feature importance defining marker genes segregating the 13 cell types can be found in Figure29BGenerating Tumor Hybrid Scores and assigning leukemia cells to hybrid populations

[0287] Tumor Hybrid gene signatures were generated as previously described16. First, normalized gene expression values were correlated to RF cell type classification probabilities along B cell progenitor cell types (HSC, Pre-B, and Immature B). Pro-B RF probability correlations were excluded; since most leukemic cells were dominantly classified as Pro-B with secondary classifications along B cell lineage cell types, genes that highly correlated to Pro-B RF probabilities were not Pro-B-specific. To ensure that genes in each hybrid population signature were specific and unique to HSC, Pre-B, and Immature B cell types, the second- highest cell type correlation coefficient was subtracted from the highest correlation coefficientfor a given cell type. Additionally, to ensure that cell type signatures were not obfuscated by cell cycle, positive correlation values of genes with cell cycle scores were subtracted from the highest correlation coefficient of a given cell type. After performing these corrections, the top 30 correlated genes to HSC, Pre-B, and Immature B cell types were included in their respective hybrid gene signatures; a threshold of 30 genes was selected based on the approximate elbow in corrected correlation values for each hybrid signature. Likewise, Pro-B gene scores were defined by the top 30 differentially expressed genes in healthy Pro-B cells (Figures S6A;Table S5)

[0288] Tumor cells were scored by these HSC, Pro-B, Pre-B, and Immature B gene signatures using the Seurat v4 AddModule Score function, and consequently assigned to hybrid populations similarly to what has been described previously.A60ASingle cells were classified into HSC-like, PreB-like, and Immature-like hybrid populations based on their highest hybrid cell type signature score, which Applicants required to be > 0.5 + that cell's Pro-B score. All other cells were classified as Pro-B like cells, which were characterized by strong Pro-B gene expression and weak or no co-expression of other cell type hybrid signature scores. The classifications based on these hybrid score distributions and relative to their B cell lineage RF prediction probabilities is demonstrated in Figure 30B.Mutual information of transcription factor activities with tumor hybrids

[0289] Applicants sought to elucidate gene programs whose activity associated with the tumor hybrid populations defined above. Given the highly entropic co-expression of tumor hybrid signatures with Pro-B marker genes, Applicants utilized mutual information as a metric for the potentially non-linear mutual dependence of gene expression with hybrid-defined developmental marker genes. Within respective hybrid subpopulations of each individual PDX line's pre-treatment and progression time points, Applicants calculated the average normalized mutual information (NMI) of all highly expressed genes across the top 30 genes in each hybrid population signature, using raw gene counts as input. Within each PDX sample and hybrid population, MI values between each gene-gene pair were generated using R infotheo package mutinformation function with the Miller-Madow asymptotic bias corrected empirical estimator and normalized to scale values between 0 and 1 as a relative, comparable metric between samples83. Applicants interpret these NMI values as a metric for genes whose expression relatively scale with hybrid population identity.

[0290] To identify cooperatively expressed genes that are collectively mutually informed with tumor hybrid signatures, Applicants utilized the collectRI transcription factor accessibility database along with the decoupleR package to in silico predict mutually informed transcription factor (TF) activity with tumor hybrid identity84. Averaged NMI values for each PDX sample hybrid were used as input with the run ulm function to estimate the linear relationship between TF-target genes and their hybrid marker gene expression. Within each PDX samples, significant TFs were ordered by their variance in mutually informed activity between hybrid populations, and the top 30 of these TFs were selected for further inspection of scaled predicted activity between hybrid subsets. NMI values and in silico predicted TF activities for each healthy reference population (HSC for HSC-hyb, Pre-BI and Pre-BII for PreB-hyb, Immature B for ImmB-hyb) were generated analogously and post-hoc compared to their leukemic hybrid counterparts (subset shown in Figure 30D), demonstrating that the majority of leukemic hybrid-defining TF activities were conserved with their healthy counterparts, with a couple of TFs.Defining developmental skews in Smart-seq2 PDX samples

[0291] Given the paucity of RF-classified immature B cells in the SS2 leukemic dataset (Figure 32A), Applicants identified genes that were Pearson correlated with Pre-B RF probabilities and with progenitor population (HSC, GMP, Pro-Mono, Early-Erythroid) probabilities. Applicants found that genes correlated with progenitor RF probabilities negatively correlated with Pre-B RF probabilities in leukemic cells and vice versa, enabling us to define a spectrum of differentiation between progenitor and later-stage B cell developmental stages (Figures 32C & 32D). Progenitor-like and PreB-like scores were generated by scoring leukemic cells over the top 30 genes significantly correlated to their respective RF probabilities (Table S7). Each cell's location on the leukemic differentiation spectrum was defined by its (PreB-like score - Progenitor-like score).Identifying somatic variants in full-length Smart-seq2 (SS2) scRNA-seq libraries

[0292] Each sample's SS2 FASTQ files were aligned to hgl9 using STAR (version 2.6.0c) and then sorted and indexed with SAMtools (version 1.13)85’86. 16 genomic loci, nominated based on recurrently identified SNVs from bulk RNA-seq in the genes KRAS, NRAS, PTPN11, GNB1, ABL1, and STAT5A (Figure 33A), were assessed for wild-type or mutant transcript detection by a custom script utilizing the Pysam library (version 0.16.0.1)87In particular, for each locus of interest, each cell was marked as "NC" if there was no coverage atthe locus, marked with 0 if all overlapping reads matched the reference allele, or marked as mutant if there were overlapping reads that did not match the reference allele.Predicting chromosomal number variations (CNVs) in SS2 scRNA-seq libraries with inferCNV

[0293] To identify SS2 leukemic cells harboring CNVs and in silico elucidate subclonal heterogeneity within tumors, Applicants estimated single-cell CNVs as previously described by computing the average expression in a sliding window of 100 genes within each chromosome after sorting the detected genes by their hgl9 genome-defined chromosomal coordinates88’89. Applicants used all healthy bone marrow SS2 cells identified above (Figure 32B) as reference normal populations for this analysis. Complete information on the inferCNV workflow used for this analysis can be found here: github.com / broadinstitute / inf erCNV / wiki, using baseline input parameters for SS2 data and for the i6 HMM algorithm for confident CNV- positive or negative predictions in single-cells.Module scoring single-cell transcriptomes

[0294] Module scores of all gene signatures over single-cells were annotated using the Seurat v4 AddModule Score function, which calculates the average expression levels of genes in a gene list relative to all other genes with comparable normalized gene expression. Quiescent cells were binned based on positive scores for a literature-derived quiescence gene signature derived from human hematopoietic cells90. Applicants utilized previously established signatures for Gl / S (n=43 genes) and G2 / M (n=55 genes) to place each cell along this dynamic process89; after inspecting the distribution of scores in the complete dataset, Applicants considered any cell > 1.5 SD above the mean for either the Gl / S or the G2 / M scores to be cycling16. Senescence scores were derived from the top 50 genes significantly differentially expressed in the SS2 DFAB-25157 RAS-mutant cells in

[0295] remission compared to all other RAS-mutant SS2 cells (Figure 22F; Table 7).Defining stress-autophagy, pre-BCR signaling, and inflammation transcriptional programs at remission

[0296] To define heterogeneous, correlated transcriptional states defining PDX tumors that emerge in MRD, Applicants first performed differential gene expression analysis between paired pre-treatment and MRD cells within the same PDX line to identify genes that significantly increase expression at remission. A total of 40 MRD state-defining genes were identified based on significant upregulation in at least two PDX-specific MRD differentiallyexpressed gene (DEG) lists. Performing gene-gene Pearson correlation across the expression of these 40 shared MRD-high DEGs in all remission leukemic cells revealed three correlated modules of genes. To expand these three modules, Applicants identified the top 30 genes significantly correlated (>2 standard deviations above median Pearson correlation) with the top differentially expressed gene in each module (Table 8). Pathway enrichment of significantly correlated genes was performed over msigDB Reactome gene sets for functional annotation, and to nominate targeted inhibitors of state (Figures 34C).REFERENCES1. Byrgazov, K., Lucini, C.B., Valent, P., Hantschel, O., and Lion, T. (2018). BCR- ABL1 compound mutants display differential and dose-dependent responses to ponatinib. Haematologica 103, el0-el2.2. Soverini, S., De Benedittis, C., Papayannidis, C., Paolini, S., Venturi, C., lacobucci, I., Luppi, M., Bresciani, P., Salvucci, M., Russo, D., et al. (2014). Drug resistance and BCR-ABL kinase domain mutations in Philadelphia chromosome-positive acute lymphoblastic leukemia from the imatinib to the second-generation tyrosine kinase inhibitor era: The main changes are in the type of mutations, but not in the frequency of mutation involvement. Cancer 120, 1002-1009.3. Eide, C.A., Zabriskie, M.S., Savage Stevens, S.L., Antelope, O., Vellore, N.A., Than, EL, Schultz, A.R., Clair, P., Bowler, A.D., Pomicter, A.D., et al. (2019). Combining the Allosteric Inhibitor Asciminib with Ponatinib Suppresses Emergence of and Restores Efficacy against Highly Resistant BCR-ABL1 Mutants. Cancer Cell 36, 431- 443. e5.4. Bhang, H.C., Ruddy, D.A., Krishnamurthy Radhakrishna, V., Caushi, J.X., Zhao, R., Hirns, M.M., Singh, A.P., Kao, I., Rakiec, D., Shaw, P., et al. (2015). Studying clonal dynamics in response to cancer therapy using high-complexity barcoding. Nat Med 21, 440-448.5. Hochhaus, A., Kreil, S., Corbin, A.S., La Rosee, P., Muller, M.C., Lahaye, T., Hanfstein, B., Schoch, C., Cross, N.C.P., Berger, U., et al. (2002). Molecular and chromosomal mechanisms of resistance to imatinib (STI571) therapy. Leukemia 16, 2190-2196.6. Good, Z., Sarno, J., Jager, A., Samusik, N., Aghaeepour, N., Simonds, E.F., White, L., Lacayo, N.J., Fantl, W.J., Fazio, G., et al. (2018). Single-cell developmental classification of B cell precursor acute lymphoblastic leukemia at diagnosis reveals predictors of relapse. Nat Med 24, 474-483.7. Huang, X., Li, Y., Zhang, J., Yan, L., Zhao, H., Ding, L., Bhatara, S., Yang, X., Yoshimura, S., Yang, W., et al. (2024). Single-cell systems pharmacology identifies development-driven drug response and combination therapy in B cell acute lymphoblastic leukemia. Cancer Cell 42, 552-567. e6.8. Kim, J.C., Chan-Seng-Yue, M., Ge, S., Zeng, A.G.X., Ng, K., Gan, O.I., Garcia-Prat, L., Flores-Figueroa, E., Woo, T., Zhang, A.X.W., et al. (2023). Transcriptomic classes of BCR-ABL1 lymphoblastic leukemia. Nat Genet 55, 1186-1197.Bastian, L., Beder, T., Barz, M.J., Bendig, S., Bartsch, L., Walter, W., Wolgast, N., Brandl, B., Rohrandt, C., Hansen, B.-T., et al. (2024). Developmental trajectories and cooperating genomic events define molecular subtypes of BCR:: ABL1 -positive ALL. Blood 143, 1391-1398. Chan, L.N., Murakami, M.A., Robinson, M.E., Caeser, R., Sadras, T., Lee, J., Cosgun, K.N., Kume, K., Khairnar, V., Xiao, G., et al. (2020). Signalling input from divergent pathways subverts B cell transformation. Nature 583, 845-851. Ebinger, S., Ozdemir, E.Z., Ziegenhain, C., Tiedt, S., Castro Alves, C., Grunert, M., Dworzak, M., Lutz, C., Turati, V.A., Enver, T., et al. (2016). Characterization of Rare, Dormant, and Therapy-Resistant Cells in Acute Lymphoblastic Leukemia. Cancer Cell 30, 849-862. Hurtz, C., Wertheim, G.B., Loftus, J.P., Blumenthal, D., Lehman, A., Li, Y., Bagashev, A., Manning, B., Cummins, K.D., Burkhardt, J.K., et al. (2020). Oncogene-independent BCR-like signaling adaptation confers drug resistance in Ph- like ALL. Journal of Clinical Investigation 130, 3637-3653. Giustacchini, A., Thongjuea, S., Barkas, N., Woll, P.S., Povinelli, B.J., Booth, C.A.G., Sopp, P., Norfo, R., Rodriguez-Meira, A., Ashley, N., et al. (2017). Singlecell transcriptomics uncovers distinct molecular signatures of stem cells in chronic myeloid leukemia. Nat Med 23, 692-702. Eadie, L.N., Saunders, V.A., Branford, S., White, D.L., and Hughes, T.P. (2018). The new allosteric inhibitor asciminib is susceptible to resistance mediated by ABCB1 and ABCG2 overexpression in vitro. Oncotarget 9, 13423-13437. Fennell, K.A., Vassiliadis, D., Lam, E.Y.N., Martelotto, L.G., Balic, J. J., Hollizeck, S., Weber, T.S., Semple, T., Wang, Q., Miles, D.C., et al. (2022). Non-genetic determinants of malignant clonal fitness at single-cell resolution. Nature 601, 125— 131. van Galen, P., Hovestadt, V., Wadsworth II, M.H., Hughes, T.K., Griffin, G.K., Battaglia, S., Verga, J. A., Stephansky, J., Pastika, T.J., Lombardi Story, J., et al. (2019). Single-Cell RNA-Seq Reveals AML Hierarchies Relevant to Disease Progression and Immunity. Cell 176, 1265-1281. e24. Velten, L., Story, B.A., Hernandez-Malmierca, P., Raffel, S., Leonce, D.R., Milbank, J., Paulsen, M., Demir, A., Szu-Tu, C., Frbmel, R., et al. (2021). Identification of leukemic and pre-leukemic stem cells by clonal tracking from single-cell transcriptomics. Nat Commun 12, 1366. Nam, A.S., Kim, K.-T., Chaligne, R., Izzo, F., Ang, C., Taylor, J., Myers, R.M., Abu- Zeinah, G., Brand, R., Omans, N.D., et al. (2019). Somatic mutations and cell identity linked by Genotyping of Transcriptomes. Nature 571, 355-360. Li, Q., Bohin, N., Wen, T., Ng, V., Magee, J., Chen, S.-C., Shannon, K., and Morrison, S.J. (2013). Oncogenic Nras has bimodal effects on stem cells that sustainably increase competitiveness. Nature 504, 143-147. Robinson, T.M., Bowman, R.L., Persaud, S., Liu, Y., Neigenfind, R., Gao, Q., Zhang, J., Sun, X., Miles, L.A., Cai, S.F., et al. (2023). Single-cell genotypic and phenotypic analysis of measurable residual disease in acute myeloid leukemia. Sci Adv 9, eadg0488. Luskin, M.R., Murakami, M.A., Manalis, S.R., and Weinstock, D.M. (2018). Targeting minimal residual disease: a path to cure? Nat Rev Cancer 18, 255-263. Tettero, J.M., Ngai, L.L., Bachas, C., Breems, D.A., Fischer, T., Gjertsen, B.T., Gradowska, P., Griskevicius, L., Janssen, J.J.W.M., Juliusson, G., et al. (2023).Measurable residual disease-guided therapy in intermediate-risk acute myeloid leukemia patients is a valuable strategy in reducing allogeneic transplantation without negatively affecting survival. Haematologica 108, 2794-2798. Dekker, S.E., Rea, D., Cayuela, J.-M., Arnhardt, I., Leonard, J., and Heuser, M. (2023). Using Measurable Residual Disease to Optimize Management of AML, ALL, and Chronic Myeloid Leukemia. Am Soc Clin Oncol Educ Book 43, e390010. Wieduwilt, M.J. (2022). Ph+ ALL in 2022: is there an optimal approach? Hematology Am Soc Hematol Educ Program 2022, 206-212. Agudo, J., Aguirre-Ghiso, J. A., Bhatia, M., Chodosh, L.A., Correia, A.L., and Klein, C.A. (2024). Targeting cancer cell dormancy. Nat Rev Cancer 24, 97-104. Wylie, A.A., Schoepfer, J., Jahnke, W., Cowan-Jacob, S.W., Loo, A., Furet, P., Marzinzik, A.L., Pelle, X., Donovan, J., Zhu, W., et al. (2017). The allosteric inhibitor ABL001 enables dual targeting of BCR-ABL1. Nature 543, 733-737. Townsend, E.C., Murakami, M.A., Christodoulou, A., Christie, A.L., Koster, J., DeSouza, T.A., Morgan, E.A., Kallgren, S.P., Liu, H., Wu, S.-C., et al. (2016). The Public Repository of Xenografts Enables Discovery and Randomized Phase Il-like Trials in Mice. Cancer Cell 29, 574-586. Gokbuget, N., Stanze, D., Beck, J., Diedrich, H., Horst, H.-A., Huttmann, A., Kobbe, G., Kreuzer, K.-A., Leimer, L., Reichle, A., et al. (2012). Outcome of relapsed adult lymphoblastic leukemia depends on response to salvage chemotherapy, prognostic factors, and performance of stem cell transplantation. Blood 120, 2032-2041. Martinelli, G., lacobucci, I., Storlazzi, C.T., Vignetti, M., Paoloni, F., Cilloni, D., Soverini, S., Vitale, A., Chiaretti, S., Cimino, G., et al. (2009). IKZF1 (Ikaros) deletions in BCR-ABL1 -positive acute lymphoblastic leukemia are associated with short disease-free survival and high rate of cumulative incidence of relapse: a GIMEMA AL WP report. J Clin Oncol 27, 5202-5207. Pfeifer, H., Wassmann, B., Pavlova, A., Wunderle, L., Oldenburg, J., Binckebanck, A., Lange, T., Hochhaus, A., Wystub, S., Bruck, P., et al. (2007). Kinase domain mutations of BCR-ABL frequently precede imatinib-based therapy and give rise to relapse in patients with de novo Philadelphia-positive acute lymphoblastic leukemia (Ph+ ALL). Blood 110, 727-734. Hughes, T.K., Wadsworth, M.H., Gierahn, T.M., Do, T., Weiss, D., Andrade, P.R., Ma, F., de Andrade Silva, B.J., Shao, S., Tsoi, L.C., et al. (2020). Second-Strand Synthesis-Based Massively Parallel scRNA-Seq Reveals Cellular States and Molecular Features of Human Inflammatory Skin Pathologies. Immunity 53, 878- 894. e7. Baccin, C., Al-Sabah, J., Velten, L., Helbling, P.M., Griinschlager, F., Hemandez- Malmierca, P., Nomb el a- Arrieta, C., Steinmetz, L.M., Trumpp, A., and Haas, S. (2020). Combined single-cell and spatial transcriptomics reveal the molecular, cellular and spatial bone marrow niche organization. Nat Cell Biol 22, 38-48. Slobodnyuk, K., Radic, N., Ivanova, S., Llado, A., Trempolec, N., Zorzano, A., and Nebreda, A.R. (2019). Autophagy -induced senescence is regulated by p38a signaling. Cell Death Dis 10, 376. Naka, K., Jomen, Y., Ishihara, K., Kim, J., Ishimoto, T., Bae, E.-J., Mohney, R.P., Stirdivant, S.M., Oshima, H., Oshima, M., et al. (2015). Dipeptide species regulate p38MAPK-Smad3 signalling to maintain chronic myelogenous leukaemia stem cells. Nat Commun 6, 8039.Sui, X., Kong, N., Ye, L., Han, W., Zhou, J., Zhang, Q., He, C., and Pan, H. (2014). p38 and JNK MAPK pathways control the balance of apoptosis and autophagy in response to chemotherapeutic agents. Cancer Lett 344, 174-179. Naeim, F., Song, S.X., Rao, P.N., and Phan, R.T. (2018). Atlas of Hematopathology: Morphology, Immunophenotype, Cytogenetics, and Molecular Approaches 2nd ed. (Elsevier, Inc). Burg, T.P., Godin, M., Knudsen, S.M., Shen, W., Carlson, G., Foster, J.S., Babcock, K., and Manalis, S.R. (2007). Weighing of biomolecules, single cells and single nanoparticles in fluid. Nature 446, 1066-1069. Feijo Delgado, F., Cermak, N., Hecht, V.C., Son, S., Li, Y., Knudsen, S.M., Oleum, S., Higgins, J.M., Chen, J., Grover, W.H., et al. (2013). Intracellular water exchange for measuring the dry mass, water mass and changes in chemical composition of living cells. PLoS One 8, e67590. Hecht, V.C., Sullivan, L.B., Kimmerling, R.J., Kim, D.-H., Hosios, A.M., Stockslager, M.A., Stevens, M.M., Kang, J.H., Wirtz, D., Vander Heiden, M.G., et al. (2016). Biophysical changes reduce energetic demand in growth factor-deprived lymphocytes. J Cell Biol 212, 439-447. Cermak, N., Oleum, S., Delgado, F.F., Wasserman, S.C., Payer, K.R., A Murakami, M., Knudsen, S.M., Kimmerling, R.J., Stevens, M.M., Kikuchi, Y., et al. (2016). High-throughput measurement of single-cell growth rates using serial microfluidic mass sensor arrays. Nat Biotechnol 34, 1052-1059. Kimmerling, R.J., Prakadan, S.M., Gupta, A. J., Calistri, N.L., Stevens, M.M., Oleum, S., Cermak, N., Drake, R.S., Pelton, K., De Smet, F., et al. (2018). Linking single-cell measurements of mass, growth rate, and gene expression. Genome Biology 19, 207. Neurohr, G.E., Terry, R.L., Lengefeld, J., Bonney, M., Brittingham, G.P., Moretto, F., Miettinen, T.P., Vaites, L.P., Soares, L.M., Paulo, J.A., et al. (2019). Excessive Cell Growth Causes Cytoplasm Dilution And Contributes to Senescence. Cell 176, 1083- 1097. el8. Son, S., Tzur, A., Weng, Y., Jorgensen, P., Kim, J., Kirschner, M.W., and Manalis, S.R. (2012). Direct observation of mammalian cell growth and size regulation. Nat Methods 9, 910-912. Miettinen, T.P., Kang, J.H., Yang, L.F., and Manalis, S.R. (2019). Mammalian cell growth dynamics in mitosis. Elife 8, e44700. Marquart, J., Chen, E.Y., and Prasad, V. (2018). Estimation of the Percentage of US Patients With Cancer Who Benefit From Genome-Driven Oncology. JAMA Oncol 4, 1093-1098. Letai, A. (2017). Functional precision cancer medicine — moving beyond pure genomics. Nat Med 23, 1028-1035. Letai, A., Bhola, P., and Welm, A.L. (2022). Functional precision oncology: Testing tumors with drugs to identify vulnerabilities and novel combinations. Cancer Cell 40, 26-35. Braun, T.P., Eide, C.A., and Druker, B.J. (2020). Response and Resistance to BCR- ABLl-Targeted Therapies. Cancer Cell 37, 530-542. Boumahdi, S., and de Sauvage, F.J. (2020). The great escape: tumour cell plasticity in resistance to targeted therapy. Nat Rev Drug Discov 19, 39-56. Duy, C., Li, M., Teater, M., Meydan, C., Garrett-Bakelman, F.E., Lee, T.C., Chin, C.R., Durmaz, C., Kawabata, K.C., Dhimolea, E., et al. (2021). ChemotherapyInduces Senescence-Like Resilient Cells Capable of Initiating AML Recurrence. Cancer Discov 11, 1542-1561. Aggarwal, R., Huang, J., Alumkal, J. J., Zhang, L., Feng, F.Y., Thomas, G.V., Weinstein, A.S., Friedl, V., Zhang, C., Witte, O.N., et al. (2018). Clinical and Genomic Characterization of Treatment-Emergent Small-Cell Neuroendocrine Prostate Cancer: A Multi-institutional Prospective Study. JCO 36, 2492-2503. Uruga, H., Fujii, T., Nakamura, N., Moriguchi, S., Kishi, K., and Takaya, H. (2020). Squamous cell transformation as a mechanism of acquired resistance to tyrosine kinase inhibitor in EGFR-mutated lung adenocarcinoma: a report of two cases. Respirology Case Reports 8, e00521. Sequist, L.V., Waltman, B.A., Dias-Santagata, D., Digumarthy, S., Turke, A.B., Fidias, P., Bergethon, K., Shaw, A.T., Gettinger, S., Cosper, A.K., et al. (2011). Genotypic and Histological Evolution of Lung Cancers Acquiring Resistance to EGFR Inhibitors. Science Translational Medicine 3, 75ra26-75ra26. Wang, Y.-C., Morrison, G., Gillihan, R., Guo, J., Ward, R.M., Fu, X., Botero, M.F., Healy, N.A., Hilsenbeck, S.G., Phillips, G.L., et al. (2011). Different mechanisms for resistance to trastuzumab versus lapatinib in HER2-positive breast cancers— role of estrogen receptor and HER2 reactivation. Breast Cancer Res 13, R121. Luskin, M.R., Murakami, M.A., Keating, J., Winer, E.S., Garcia, J.S., Stahl, M., Wadleigh, M., Flamand, Y., Neuberg, D.S., Galinsky, I., et al. (2023). A Phase I Study of Asciminib (ABL001) in Combination with Dasatinib and Prednisone for BCR-ABL1 -Positive ALL and Blast Phase CML in Adults. Blood 142, 965. Keating, A.K., Gossai, N., Phillips, C.L., Maloney, K., Campbell, K., Doan, A., Bhojwani, D., Burke, M.J., and Verneris, M.R. (2019). Reducing minimal residual disease with blinatumomab prior to HCT for pediatric patients with acute lymphoblastic leukemia. Blood Advances 3, 1926-1929. Gokbuget, N., Zugmaier, G., Dombret, H., Stein, A., Bonifacio, M., Graux, C., Faul, C., Briiggemann, M., Taylor, K., Mergen, N., et al. (2020). Curative outcomes following blinatumomab in adults with minimal residual disease B-cell precursor acute lymphoblastic leukemia. Leukemia & Lymphoma 61, 2665-2673. Salek, M., Li, N., Chou, H.-P., Saini, K., Jovic, A., Jacobs, K.B., Johnson, C., Lu, V., Lee, E.J., Chang, C., et al. (2023). COSMOS: a platform for real-time morphologybased, label-free cell sorting using deep learning. Commun Biol 6, 971. Chan, J.M., Zaidi, S., Love, J.R., Zhao, J.L., Setty, M., Wadosky, K.M., Gopalan, A., Choo, Z.-N., Persad, S., Choi, J., et al. (2022). Lineage plasticity in prostate cancer depends on JAK / STAT inflammatory signaling. Science 377, 1180-1191. Raghavan, S., Winter, P.S., Navia, A.W., Williams, H.L., DenAdel, A., Lowder, K.E., Galvez-Reyes, J., Kalekar, R.L., Mulugeta, N., Kapner, K.S., et al. (2021). Microenvironment drives cell state, plasticity, and drug response in pancreatic cancer. Cell 184, 6119-6137. e26. Rambow, F., Rogiers, A., Marin-Bejar, O., Aibar, S., Femel, J., Dewaele, M., Karras, P., Brown, D., Chang, Y.H., Debiec-Rychter, M., et al. (2018). Toward Minimal Residual Disease-Directed Therapy in Melanoma. Cell 174, 843-855. el9. Hahn, W.C., Bader, J.S., Braun, T.P., Califano, A., Clemons, P.A., Druker, B.J., Ewald, A.J., Fu, H., Jagu, S., Kemp, C.J., et al. (2021). An expanded universe of cancer targets. Cell 184, 1142-1155.Odejide, O., Weigert, O., Lane, A. A., Toscano, D., Lunning, M.A., Kopp, N., Kim, S., van Bodegom, D., Bolla, S., Schatz, J.H., et al. (2014). A targeted mutational landscape of angioimmunoblastic T-cell lymphoma. Blood 123, 1293-1296. Kluk, M.J., Lindsley, R.C., Aster, J.C., Lindeman, N.I., Szeto, D., Hall, D., and Kuo, F.C. (2016). Validation and Implementation of a Custom Next-Generation Sequencing Clinical Assay for Hematologic Malignancies. The Journal of Molecular Diagnostics 18, 507-515. Stockslager, M.A., Malinowski, S., Touat, M., Yoon, J.C., Geduldig, J., Mirza, M., Kim, A.S., Wen, P.Y., Chow, K.-H., Ligon, K.L., et al. (2021). Functional drug susceptibility testing using single-cell mass predicts treatment outcome in patient- derived cancer neurosphere models. Cell Reports 37, 109788. Trombetta, J. J., Gennert, D., Lu, D., Satija, R., Shalek, A.K., and Regev, A. (2014). Preparation of Single-Cell RNA-Seq Libraries for Next Generation Sequencing. Current Protocols in Molecular Biology 107, 4.22.1-4.22.17. Themeau, T. (2024). A Package for Survival Analysis in R. Li, H., and Durbin, R. (2009). Fast and accurate short read alignment with Burrows- Wheeler transform. Bioinformatics 25, 1754-1760. McKenna, A., Hanna, M., Banks, E., Sivachenko, A., Cibulskis, K., Kemytsky, A., Garimella, K., Altshuler, D., Gabriel, S., Daly, M., et al. (2010). The Genome Analysis Toolkit: A MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 20, 1297-1303. DePristo, M.A., Banks, E., Poplin, R., Garimella, K.V., Maguire, J.R., Hartl, C., Philippakis, A. A., del Angel, G., Rivas, M.A., Hanna, M., et al. (2011). A framework for variation discovery and genotyping using next-generation DNA sequencing data. Nat Genet 43, 491-498. Cibulskis, K., Lawrence, M.S., Carter, S.L., Sivachenko, A., Jaffe, D., Sougnez, C., Gabriel, S., Meyerson, M., Lander, E.S., and Getz, G. (2013). Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples. Nat Biotechnol 31, 213-219. McLaren, W., Pritchard, B., Rios, D., Chen, Y., Flicek, P., and Cunningham, F. (2010). Deriving the consequences of genomic variants with the Ensembl API and SNP Effect Predictor. Bioinformatics 26, 2069-2070. Bi, W.L., Greenwald, N.F., Ramkissoon, S.H., Abedalthagafi, M., Coy, S.M., Ligon, K.L., Mei, Y., MacConaill, L., Ducar, M., Min, L., et al. (2017). Clinical Identification of Oncogenic Drivers and Copy -Number Alterations in Pituitary Tumors. Endocrinology 158, 2284-2291. Li, B., Gould, J., Yang, Y., Sarkizova, S., Tabaka, M., Ashenberg, O., Rosen, Y., Slyper, M., Kowalczyk, M.S., Villani, A.-C., et al. (2020). Cumulus provides cloudbased data analysis for large-scale single-cell and single-nucleus RNA-seq. Nat Methods 17, 793-798. Butler, A., Hoffman, P., Smibert, P., Papalexi, E., and Satija, R. (2018). Integrating single-cell transcriptomic data across different conditions, technologies, and species. Nat Biotechnol 36, 411-420. Weinreb, C., Wolock, S., and Klein, A.M. (2018). SPRING: a kinetic interface for visualizing high dimensional single-cell expression data. Bioinformatics 34, 1246- 1248.Kotliar, D., Veres, A., Nagy, M.A., Tabrizi, S., Hodis, E., Melton, D.A., and Sabeti, P.C. (2019). Identifying gene expression programs of cell-type identity and cellular activity with single-cell RNA-Seq. eLife 8, e43803. Gavish, A., Tyler, M., Greenwald, A.C., Hoefflin, R., Simkin, D., Tschemichovsky, R., Galili Darnell, N., Somech, E., Barbolin, C., Antman, T., et al. (2023). Hallmarks of transcriptional intratumour heterogeneity across a thousand tumours. Nature 618, 598-606. Kazer, S.W., Aicher, T.P., Muema, D.M., Carroll, S.L., Ordovas-Montanes, J., Miao, V.N., Tu, A.A., Ziegler, C.G.K., Nyquist, S.K., Wong, E.B., et al. (2020). Integrated single-cell analysis of multicellular immune dynamics during hyperacute HIV-1 infection. Nat Med 26, 511-518. Kuhn, M. (2008). Building Predictive Models in R Using the caret Package. Journal of Statistical Software 28, 1-26. Wright, M.N., and Ziegler, A. (2017). ranger: A Fast Implementation of Random Forests for High Dimensional Data in C++ and R. Journal of Statistical Software 77, 1-17. Janitza, S., Celik, E., and Boulesteix, A.-L. (2018). A computationally fast variable importance test for random forests for high-dimensional data. Advances in Data Analysis and Classification 12, 885-915. Meyer, P.E. (2008). Information-Theoretic Variable Selection and Network Inference from Microarray Data. In. Badia-i-Mompel, P., Velez Santiago, J., Braunger, J., Geiss, C., Dimitrov, D., Miiller- Dott, S., Taus, P., Dugourd, A., Holland, C.H., Ramirez Flores, R.O., et al. (2022). decoupleR: ensemble of computational methods to infer biological activities from omics data. Bioinformatics Advances 2, vbac016. Dobin, A., Davis, C.A., Schlesinger, F., Drenkow, J., Zaleski, C., Jha, S., Batut, P., Chaisson, M., and Gingeras, T.R. (2013). STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15-21. Danecek, P., Bonfield, J.K., Liddle, J., Marshall, J., Ohan, V., Pollard, M.O., Whitwham, A., Keane, T., McCarthy, S.A., Davies, R.M., et al. (2021). Twelve years of SAMtools and BCFtools. GigaScience 10, giab008. Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., Marth, G., Abecasis, G., Durbin, R., and 1000 Genome Project Data Processing Subgroup (2009). The Sequence Alignment / Map format and SAMtools. Bioinformatics 25, 2078-2079. Patel, A.P., Tirosh, I., Trombetta, J.J., Shalek, A.K., Gillespie, S.M., Wakimoto, H., Cahill, D.P., Nahed, B.V., Curry, W.T., Martuza, R.L., et al. (2014). Single-cell RNA-seq highlights intratumoral heterogeneity in primary glioblastoma. Science 344, 1396-1401. Tirosh, I., Izar, B., Prakadan, S.M., Wadsworth, M.H., Treacy, D., Trombetta, J.J., Rotem, A., Rodman, C., Lian, C., Murphy, G., et al. (2016). Dissecting the multicellular ecosystem of metastatic melanoma by single-cell RNA-seq. Science 352, 189-196. Graham, S.M., Vass, J.K., Holyoake, T.L., and Graham, G.J. (2007). Transcriptional Analysis of Quiescent and Proliferating CD34+ Human Hemopoietic Cells from Normal and Chronic Myeloid Leukemia Sources. Stem Cells 25, 3111-3120.

[0297] Table 1. Clinical characteristics of patients whose tumors were used to generate PDX models.Hyperdiploid

[0298] Table 2. Clinical characteristics of patients whose tumors were used to generate PDX models.

[0299] Table 3. cNMF meta-GEP gene lists.

[0300] Table 4. Seq-Well derived Tumor Hybrid signatures.

[0301] Table 5. Patient characteristics and clinical trial outcomes.

[0302] Table 6. SS2-derived Tumor Hybrid signatures.

[0303] Table 7. Senescence-Like signature.

[0304] Table 8. MRD State signatures.

[0305] Table 9. Flow cytometry antibodies.***

[0306] Various modifications and variations of the described methods, pharmaceutical compositions, and kits of the disclosure will be apparent to those skilled in the art without departing from the scope and spirit of the disclosure. Although the disclosure has been described in connection with specific embodiments, it will be understood that it can be further modified and that the disclosure as claimed should not be unduly limited to such particular embodiments. Indeed, various modifications of the described modes for carrying out the disclosure that are obvious to those skilled in the art are intended to be within the scope of the disclosure. This application is designed to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure that come within known customary practice within the art to which the disclosure pertains and may be applied to the essential features herein before set forth.

[0307] References and citations to other documents, such as patents, patent applications, patent publications, journals, books, papers, and web contents, have been made in this disclosure. All such documents are hereby incorporated herein by reference in their entirety for all purposes. Any material, or portion thereof, that is said to be incorporated by reference herein but which conflicts with existing definitions, statements, or other disclosure material explicitly set forth herein is only incorporated to the extent that no conflict arises between that incorporated material and the present disclosure material. In the event of a conflict, the conflict is to be resolved in favor of the present disclosure as the preferred disclosure.

Claims

CLAIMSWhat is claimed is:

1. A method of treating leukemia comprising: detecting in a sample from a subject suffering from leukemia a cell state, wherein the cell state comprises a Pre-B cell receptor (Pre-BCR) state or a stress / autophagy state; wherein if a pre-BCR state is detected, then administering to the subject a treatment comprising (i) a BCR signaling pathway inhibitor, or (ii) a tyrosine kinase inhibitor (TKI) inhibitor and a SYK inhibitor and if a stress-autophagy state is detected, administering a TKI inhibitor and a p38 inhibitor to the subject.

2. The method of claim 1, wherein the leukemia is an oncogene-addicted leukemia.

3. The method of claim 1, wherein the leukemia is acute lymphoblastic leukemia (ALL).

4. The method of claim 2, wherein the oncogene-addicted leukemia is BCR-ABL1 B-cell acute lymphoblastic leukemia (B-ALL).

5. The method of claims 1 or 2, wherein the subject is in a relapse of leukemia following induction therapy.

6. The method of claim 5, wherein the induction therapy comprises treatment with a tyrosine kinase inhibitor (TKI).

7. The method of any preceding claims, wherein the cell state comprises a functional measurement of one or more phenotypes.

8. The method of any preceding claims, wherein the cell state is detected using a biophysical measurement.

9. The method of claim 8, wherein the biophysical measurement is cell mass.

10. The method of claim 9, wherein a higher cell mass indicates the stress / autophagy cell state and a lower cell mass indicates the pre-BCR state.

11. The method of claim 10, wherein the higher cell mass is between 15 to 25pg, and the lower cell mass is between 10 to 15pg.

12. The method of claim 8, wherein the biophysical measurement is size, mass, morphology, mechanical properties, or cellular stiffness.

13. The method of claim 1, wherein the TKI is selected from the group consisting of imatinib, dasatinib, nilotinib, bosutinib, and ponatinib.

14. The method of claim 1, wherein the SYK inhibitor is selected from the group consisting of fostamatinib, entospletinib, TAK-659, R406, cerdulatinib, BAY 61-3606.

15. The method of claim 1, wherein the p38 inhibitor is selected from the group consisting of losmapimod, ruxolitinib, SB 203580, doramapimod (BIRB 796), panapimod (R- 1503).

16. The method of any one of the preceding claims, wherein the treatment is administered at a point of minimal residual disease (MRD).

17. The method of any preceding claims, where the sample from the subject comprises residual cells.

18. The method of any preceding claims further comprises determining a cycling and senescence gene expression score for the subject's sample.

19. The method of any one of the preceding claims, wherein the cycling and senescence gene expression score for the sample from the subject is categorized into one of three categories: A) cells scoring along cell cycle; B) cells poised to enter cell cycle; and C) cells scoring for senescence that have likely terminally exited the cell cycle.

20. The method of any preceding claims, wherein administering the treatment to the subject results in reduced disease burden and / or remission of the cancer, is maintained.

21. The method of any one of the preceding claims, wherein the transcriptional program comprises a Pre-B cell receptor (Pre-BCR) program comprises expression of one or more genes selected from the group consisting of IGLL1, VPREB3, TCL1A, UHRF1, H1F0, SOX11, RPS4Y1, VDAC1, CCND3, ARPP21, HMGN2, RPS26, CMTM8, IGJ, A0X2P, TMSB4X, FAM129C, SNURF, PTMA, LAT2, HIST1H3G, TP53INP1, SNRPN, NELLI, DNMT1, HMGB1, CBX1, PHGDH, LGALS9, EBF1, FLU, RPS4Y2, C16orf54, STMN1, ERGIC1, IRF2BP2, PAXS, E2F2, CD38, GNG7, POU2AF1, UBE2T, MME, NUSAP1, D0K3, PVRIG, PIP4K2A, UCP2, and DEK.

22. The method of any one of the preceding claims wherein the stress / autophagy comprises a gene selected from the group consisting of DNAJB1, HSPA1A, HSP90AB1, HSPE1, HSPH1, DNAJA1, HSPA1B, NR4A1, HSPA8, CDKN1A, HBEGF, HSP90AA1, HSPB1, HSPA6, HSPD1, ZFAND2A, TUBB4B, EIF5, UBC, EIF1, DNAJB6, DDX3X, CHORDCI, JUN, ZC3HAV1, IER3, HERPUD1, IFRD1, BAG3, SAP18, HSA1, PMAIP1, BRD2, HSPA7, MAP1LC3B, SLC2A3, ZFAS1, TCP1, EIF4G2, ATF3, INSIGI, SOD2, SAT1, DEDD2, C6orf62, CACYBP, HEXIM 1, EGR3, and CSRNP1.

23. The method of claim 4, further comprising predicting tyrosine kinase inhibitor sensitivity by: classifying cells of the BCR-ABL1 B-cell acute lymphoblastic leukemia according to B-cell developmental stage; and predicting reduced tyrosine kinase inhibitor sensitivity for leukemic cells classified as mature B-cells compared to leukemic cells classified as progenitor B-cells.

Images