Cancer Discovery

Pancreatic cancer (PC) is largely refractory to immune checkpoint blockade (ICB), although homologous recombination-deficient (HRD) tumors may derive benefit. In the POLAR trial of maintenance pembrolizumab plus olaparib after platinum-based chemotherapy for metastatic PC, responses remained heterogeneous. To define determinants of productive antitumor immunity, we integrated longitudinal blood TCR sequencing with tumor single-cell and spatial profiling. Durable benefit was associated with rare tumor-infiltrating, peripherally expanding (TIE) CD8+ T cell clonotypes, a subset of which were functionally neoantigen-reactive. TIE patients showed markedly prolonged survival beyond established genomic and immune biomarkers. Conversely, resistance was associated with spatial T cell exclusion, myCAF-rich stromal remodeling, basal-like/KRAS-associated malignant-cell programs, and expansion of CTLA4 regulatory T cells linked to local immunosuppressive remodeling. These findings define a clonotype-resolved framework for immune monitoring and identify complementary stromal, tumor-intrinsic, and regulatory immune barriers that may guide rational combination immunotherapy in PC.

Oncogenic KRAS drives >90% of pancreatic ductal adenocarcinoma (PDAC), and the pan-RAS inhibitor daraxonrasib doubles overall survival, but adaptive and acquired resistance limit the depth and duration of responses. To systematically map combination vulnerabilities, we developed RAS+, a digenic CRISPR knockout library testing 10,296 pairwise perturbations across oncogenic signaling pathways, and screened 12 cell lines, revealing genotype- and lineage-specific interactions. In KRASmut PDAC, the most differentially effective gene pair was KRAS and FRS2, an adapter coupling FGFR signaling to RAS. Combined KRAS/pan-FGFR inhibition was synergistic and cytotoxic, eradicating tumor cells and overcoming resistance in vitro and deepening and prolonging regressions in vivo. Single-cell analysis identified cancer-associated fibroblasts (CAFs) as a primary source of FGF ligands in PDAC, and CAF conditioned media or individual FGFs promoted resistance. These findings define a stroma-to-tumor adaptive FGFR circuit limiting the effects of KRAS inhibition. Thus, combined KRAS/FGFR blockade may be a strategy to improve responses.

Bulk transcriptomic classifiers stratify pancreatic ductal adenocarcinoma (PDAC) into classical and basal-like subtypes with prognostic and therapeutic relevance, yet increasing evidence indicates that these epithelial programs frequently coexist within individual tumors. How these intermediate states affect the local tumor microenvironment remains poorly defined. Here, we integrate multiplexed ion beam imaging (MIBI) with bulk RNA sequencing to resolve epithelial subtype identity at the level of spatially contiguous cancer nests and quantify their associated microenvironments. Across 47 primary tumor samples from 34 patients, we identified classical, intermediate, and basal cancer cell states at single-cell resolution and delineated discrete cancer nests with mixed or dominant subtype compositions. Distance-resolved spatial analysis reveals that basal-rich cancer nests are surrounded by locally immunosuppressive microenvironments characterized by reduced expression of MHC class II and co-stimulatory molecules in proximal myeloid cells, independent of myeloid abundance. These regions are enriched in fibroblast-dominated neighborhoods and distinct cell-cell interaction architectures. Using EcoTyper analysis of two independent bulk RNA-seq cohorts, including an OHSU discovery cohort (N = 277 patients) and TCGA as a validation cohort (N = 147 patients), we identified poor-prognosis tumor ecotypes enriched for basal epithelial states that similarly exhibited depleted myeloid antigen presentation signatures, linking spatial niche phenotypes to transcriptional ecotypes and patient outcomes. Together, these findings demonstrate that epithelial subtype programs in PDAC are organized at the level of spatially defined cancer nests and that basal cancer programs reside within localized niches of myeloid antigen presentation dysfunction, linking intratumoral architecture to immune suppression and clinical prognosis.

Relapse in cancer is frequently driven by therapy-resistant quiescent cancer stem cells. Conventional chemotherapy has been designed to target proliferating tumor cells and is generally presumed to be ineffective against non-cycling cancer stem cells. Using acute myeloid leukemia (AML) as a model, we challenge this prevailing view by showing that inhibition of the mitotic master regulator Polo-like kinase 1 (PLK1), a kinase extensively pursued for antiproliferative cancer therapy, unexpectedly eradicates quiescent leukemia stem cells (LSC) through a mechanism distinct from its canonical mitotic function. In proliferating AML cells, PLK1 inhibition (PLK1i) induced G2/M arrest and mitotic catastrophe. In contrast, quiescent LSC underwent apoptosis independent of mitotic arrest, revealing a cell-state-dependent mode of drug action. Mechanistically, PLK1i initiated a multi-step process through disruption of a previously unrecognized, stem cell-specific interaction between PLK1 and MAP1A, resulting in perturbed vesicle trafficking and endolysosomal homeostasis characterized by altered receptor internalization, vesicle accumulation and lysosomal dysfunction, ultimately culminating in apoptotic cell death. Combinatorial pharmacologic perturbation studies established microtubule regulation as a critical determinant of quiescent LSC survival, while ex vivo and in vivo assays demonstrated depletion of functionally-defined LSC following PLK1i. These findings identify a previously unrecognized role for PLK1 in intracellular trafficking and establish MAP1A-dependent control of vesicle homeostasis as a mechanistic determinant of cancer stem cell survival. More broadly, this study demonstrates that classical antimitotic compounds, including microtubule-targeting agents and PLK1 inhibitors, can eradicate both cycling leukemic blasts and quiescent LSC through distinct, cell state-dependent mechanisms, challenging proliferation-centric models of chemotherapy action.

Homologous recombination deficiency (HRD), most commonly arising from BRCA1/2 loss, is accompanied by metabolic remodeling whose contours and therapeutic implications remain poorly defined. Through pan-cancer analysis of 10,619 tumors from The Cancer Genome Atlas (TCGA) and isogenic BRCA1/2-deficient models, we show that HRD tumors display broad metabolic rewiring, with consistent alterations in glucose metabolism. Unexpectedly, BRCA1/2 loss is associated with selective upregulation of the constitutively active glycolytic enzyme Pyruvate Kinase M1 (PKM1) isoform, contrasting with the PKM2-dominant state of most cancers, and produces an elevated pyruvate kinase activity state. Overactivating PKM, with small molecules activators, including the FDA-approved drug mitapivat, triggers superoxide accumulation, disrupts PINK1-dependent mitochondrial quality control, induces DNA damage, and selectively kills BRCA2-deficient tumors in vivo, while sparing BRCA2-proficient counterparts. These findings recast HRD as a metabolic state and identify PKM activation as a candidate therapeutic strategy for biomarker-selected HR-deficient cancers.

Surgical decisions for intraductal papillary mucinous neoplasms (IPMNs), precursors of pancreatic cancer, depend on distinguishing low- from high-grade dysplasia, a distinction that molecular markers could refine. We microdissected 433 tumor regions from 135 patients and integrated transcriptomics, deep-learning histology, and KRAS/GNAS RNA-called variants. RNA programs and morphological motifs separated lesions by phenotype, intestinal versus others, rather than by dysplasia grade. Along gastric IPMN trajectories, low-grade lesions recapitulated the pit-chief polarity of normal gastric glands, whereas high-grade lesions were indistinguishable from pancreatobiliary lesions in morphology, RNA programs, and driver mutations. Intestinal IPMNs arose through two routes: a canonical GNAS-associated pathway and a KRAS-associated shift from low-grade gastric features, with lesions sharing driver mutations retaining common programs across phenotypes. Resolving these phenotypic trajectories clarified distinct progression paths and enabled phenotype-specific prediction of high-grade dysplasia in mucus aspirated from 94 resected lesions. These findings support a promising fluid-based assessment of IPMN status.

Direct KRAS inhibitors have established mutant KRAS as a clinically actionable target, yet adaptive resistance remains a major barrier to durable responses. To identify therapeutically actionable resistance mechanisms, we performed an unbiased in vivo CRISPR activation screen in an autochthonous lung adenocarcinoma model, identifying the receptor tyrosine kinase AXL as a dominant adaptive resistance driver. Pharmacologic AXL inhibition enhanced the efficacy of both allele-specific inhibition and the RAS(ON) multi-selective inhibitor daraxonrasib across lung and pancreatic cancer models, resulting in deeper and more durable suppression of MAPK signaling and improved tumor control. Beyond tumor-intrinsic effects, combined KRAS and AXL inhibition remodeled the tumor immune microenvironment, promoting an IFN{gamma}-responsive program, increased recruitment of cytotoxic T cells and sensitization to FAS-mediated apoptosis. Collectively, our findings identify AXL as a convergence point for adaptive resistance to KRAS inhibition and provide a mechanistically informed combination strategy to extend the durability of KRAS-directed therapies. Statement of SignificanceAn unbiased in vivo functional (CRISPR activation) screen identifies AXL as a convergence point for adaptive resistance to KRAS inhibition. By integrating adaptive response to KRAS inhibition with anti-tumor immunity, AXL represents a mechanistically actionable vulnerability whose inhibition deepens and prolongs responses to both allele-specific and pan-KRAS-targeted therapies.

Cancer cachexia limits treatment tolerance and survival in pancreatic ductal adenocarcinoma (PDAC), yet the tumor-derived signals driving tissue dysfunction remain poorly understood. Here, we identify serum amyloid A1 (SAA1) as a mediator of tumor-to-host communication acting through Toll-like receptor 4 (TLR4). Tumor-derived SAA1 was elevated in human PDAC and in a mouse PDAC model and disrupted both myofiber and muscle stem cell (MuSC) homeostasis. Genetic reduction of tumor-derived SAA1 uncoupled tumor progression from host wasting, preserving muscle mass and function and prolonging survival without affecting primary tumor growth. Mechanistically, SAA1-TLR4 signaling drove multicellular remodeling of the skeletal muscle microenvironment. Therapeutic TLR4 inhibition after cachexia onset restored muscle mass, function and MuSC abundance and prolonged survival independently of tumor growth. Conservation of SAA1-TLR4 signaling in human skeletal muscle identifies a therapeutically actionable tumor-host pathway and demonstrates that host deterioration can be targeted independently of tumor progression.

Loss of the Y chromosome (LOY) has been largely unexplored in pancreatic ductal adenocarcinoma (PDAC), and no mechanistic framework previously explained its role in tumor biology. We identify LOY as a recurrent, functionally consequential alteration affecting ~50% of male PDAC patients across two independent cohorts, with strong enrichment in advanced disease and, critically, in the basal subtype, establishing LOY as a clinically actionable biomarker. Comprehensive genomic profiling positions LOY as a distinct genetic event associated with TP53 mutations, implicating it in aggressive tumor evolution. Integrative analyses across bulk tumors, single-cell datasets and CRISPR-engineered PDAC models converge on epithelial-to-mesenchymal transition (EMT) as the dominant LOY-driven program. Y-chromosome deletion alone induces EMT transcriptional states, mesenchymal morphology and enhanced metastatic ability. Mechanistically, LOY disrupts epigenetic regulation through loss of Y-linked chromatin modifiers, reshaping DNA methylation at EMT- and basal-defining loci and reducing chemotherapy sensitivity. These findings establish LOY as a causal genetic driver of PDAC progression and a clinically accessible biomarker with direct therapeutic relevance.

The clinical benefit of MAPK-targeted therapies depends on greater pathway inhibition in tumors than normal tissues. Although pan-RAF inhibitors are active in RAS-mutant cancers, combining them with MEK inhibitors requires dose reductions due to toxicity, limiting efficacy. We show the toxicity results from MEK inhibitor-mediated feedback relief, which promotes RAF activation and pan-RAF inhibitor engagement in normal cells, narrowing the therapeutic index. We further demonstrate that MEK is exclusively cytosolic, and RAF/MEK glues overcome this limitation through spatial trapping. By stabilizing cytosolic RAF-MEK complexes, RAF/MEK glues prevent feedback-driven RAF activation in normal cells while maintaining inhibition of oncogenic RAF signaling in RAS-mutant tumors, where RAF is constitutively activated at the plasma membrane. Consequently, this enables full-dose combination with pan-RAF inhibitors, resulting in deeper MAPK suppression and robust tumor regressions in RAS-mutant models. Thus, by spatially controlling wild-type effectors, drug-induced proximity can be harnessed to increase tumor selectivity of pathway-targeted therapies. SignificanceMAPK-targeted therapies rarely achieve durable responses in RAS-mutant cancers due to dose-limiting toxicities. We show that RAF/MEK glues, by spatially trapping RAF, can be combined with pan-RAF inhibitors at full dose, yielding tumor-selective MAPK inhibition and tumor regressions in RAS-mutant models. Thus, drug-induced proximity can be exploited for tumor-selective therapy.

Pancreatic ductal adenocarcinoma (PDAC) exhibits profound metabolic rewiring and strong resistance to DNA-damaging therapies, yet how metabolic pathways regulate genome maintenance remains poorly understood. The hexosamine biosynthetic pathway (HBP) integrates nutrient availability with protein glycosylation through production of UDP-GlcNAc, but its role in DNA damage response (DDR) regulation is unclear. Here we show that inhibition of the HBP enzyme phosphoglucomutase-3 (PGM3) reduces DNA repair capacity in pancreatic cancer cells. Transcriptomic and functional analyses reveal that the selective PGM3 inhibitor FR054 amplifies gemcitabine-induced replication stress, disrupts ATR-CHK1 and ATM-CHK2 checkpoint signaling, and selectively impairs homologous recombination. Glycoproteomic profiling identifies the AAA+ ATPase RUVBL2 as a key metabolic-DDR node. Gemcitabine increases RUVBL2 O-GlcNAcylation, with Thr81 identified as a modified residue within the Walker A nucleotide-binding motif. Structural modelling predicts that Thr81 O-GlcNAcylation stabilizes the RUVBL1-RUVBL2 complex without compromising ATP-Mg engagement. PGM3 inhibition and Thr81 mutation similarly reduced ATR and ATM abundance and promoted persistent DNA damage, supporting a role for RUVBL2 Thr81 O-GlcNAcylation in sustaining checkpoint signalling and genome stability. Consequently, PGM3 inhibition induces a BRCAness-like state that sensitizes pancreatic cancer cells to PARP inhibition, both in vitro and in vivo, as well as to ionizing radiation. These findings reveal a nutrient-sensitive mechanism linking protein glycosylation to genome maintenance and identify HBP-dependent DNA repair as a potentially actionable vulnerability in pancreatic cancer.

Responses to macrophage-directed therapy can be transient because tumors preserve myeloid support through complementary persistence and replenishment. In esophageal squamous cell carcinoma (ESCC), CSF1R inhibition reduced established tumor-associated macrophages but was followed by expansion of Ly6C/CCR2-positive monocytic and Ly6G-positive granulocytic populations. Low-dose decitabine preferentially restricted recruited populations while sparing a LYVE1-associated macrophage state, exposing reciprocal pharmacologic blind spots. Combined treatment suppressed both arms and produced sustained control across patient-derived organoid xenograft, orthotopic, and immunocompetent models. Neutrophil depletion reproduced initial regression but not sustained control, indicating that the recruited escape arm extended beyond Ly6G-positive granulocytes. Single-cell profiling mapped these vulnerabilities onto a treatment-resolved myeloid architecture comprising a C1qa-positive TAM continuum, a C1qa-negative Ccr2/Ly6c2-high inflammatory monocytic-like compartment, and a LYVE1/MRC1-positive tissue-supportive macrophage state. Human ESCC contained corresponding macrophage programs and an adverse-outcome-associated LYVE1-rich niche. These findings identify state-aware coverage of complementary myeloid vulnerabilities as a strategy to overcome escape from macrophage-directed therapy.

Loss of the tumor suppressor LKB1 (STK11) drives metabolic reprogramming and immune evasion, but its clinical footprint is defined almost entirely by somatic mutation and deletion. Because emerging therapies aim to reactivate this pathway in structurally wild-type tumors, defining the full LKB1-deficient population (including tumors silenced by non-genomic mechanisms) is a prerequisite for patient selection. We derived a 30-gene transcriptional signature of LKB1 functional loss, training on genomically-defined STK11 loss in lung adenocarcinoma and validating in an independent squamous cohort (AUROC 0.926). It transferred to four independent non-TCGA LUAD cohorts (AUROC 0.92-0.98), outperforming the published Kaufman 16-gene classifier. Critically, restoring wild-type LKB1 in LKB1-mutant NSCLC lines reversed the signature while a kinase-dead mutant did not, establishing that it reads out LKB1 kinase function, not merely mutation status. Applied pan-cancer, the signature identified functional LKB1 loss in wild-type tumors at 7.2%, expanding total LKB1-loss prevalence 3.7-fold over genomic loss alone (2.7% [->] 9.9%), largest in esophageal, colorectal, endometrial and cutaneous cancers. This population is decoupled from LKB1 mutation and deletion frequency, yet enriched in BRAF-mutant colorectal (25% signature-positive) and HER2-amplified breast (21% signature-positive) cancers: readily-testable subgroups for screening. In lung adenocarcinoma, signature positivity rose across a co-mutation gradient from 6% background to 15% in KRAS-mutant, 35% in NRF2-pathway-mutant (KEAP1 or NFE2L2), to 64% in KRAS/NRF2 co-mutant tumors. Similar to KRAS/LKB1 co-mutant tumors, LKB1 wild-type but signature-high tumors were immune-cold, correlating negatively with antigen-presentation, interferon-{gamma} and T-cell-inflamed programs in multiple independent lung cancer cohorts. These results define a broad set of tumors with LKB1 functional loss that may open up new avenues for precision therapeutics.

Acute myeloid leukemia (AML) remains limited by high relapse rates and a scarcity of biomarker-directed therapies, underscoring the need to identify actionable vulnerabilities and mechanisms of therapeutic resistance. O6-methylguanine-DNA methyltransferase (MGMT), a DNA repair enzyme that directly reverses mutagenic O6-alkylguanine lesions, is epigenetically silenced in multiple cancers. Although MGMT silencing is predictive of temozolomide (TMZ) response in glioblastoma with significant survival advantage, prior clinical trials of TMZ in AML have shown only modest responses. Therefore, the prevalence and therapeutic relevance of MGMT silencing, as well as genetic factors that modify its therapeutic response in AML, remain to be comprehensively investigated. Here, we first profiled MGMT status across a diverse cohort of 23 de novo and 16 relapsed/refractory (R/R) primary AMLs using an integrated analysis of MGMT mRNA expression, promoter methylation, and protein expression. We found that approximately 25-30% of AMLs harbor MGMT silencing, in contrast to consistent MGMT expression in healthy CD34+ hematopoietic stem and progenitor cells. We further identified frequent loss of mismatch repair (MMR) proteins in these AML cohorts, a known resistance mechanism against TMZ in glioblastoma. Using CRISPR knockout screening and genetically engineered AML models, we found that MMR loss also drives both upfront and acquired TMZ resistance in MGMT-silenced AML. These findings identify MMR deficiency as an important and previously underappreciated genetic contributor to the modest responses observed in prior trials of TMZ in AML. To therapeutically exploit MGMT silencing while bypassing genetic determinants of resistance, including MMR deficiency, we evaluated clinically and preclinically characterized MGMT-dependent DNA-alkylating agents and identified a fluoroethylating analog of TMZ, N3-(2-fluoroethyl) imidazotetrazine (KL50), with pronounced and selective activity in MGMT-silenced AML. KL50 retained antileukemic activity irrespective of MMR status, significantly prolonging survival in humanized MISTRG6 mice harboring primary AML patient-derived xenografts. Mechanistically, KL50 induced DNA damage through the time-dependent formation of interstrand DNA crosslinks, bypassing MMR dependence and triggering a replication stress response dominated by ATR signaling. Pharmacologic ATR inhibition synergized with KL50, producing marked antileukemic activity and significantly extending survival across AML models without compromising hematologic safety. Together, these findings establish MGMT silencing as a prevalent and therapeutically actionable biomarker in AML, define MMR status as a key determinant of TMZ response in AML but not KL50 sensitivity, and provide a translational rationale for combining low-dose O6-fluoroethylating imidazotetrazines with ATR inhibitors to target both de novo and R/R AML across diverse genetic backgrounds.

The relationship between cell cycle length and differentiation competence is well established in developmental biology, particularly in embryonic stem cells, where a short G1 phase maintains pluripotency and G1 lengthening permits lineage commitment. Whether this principle operates in cancer cells, where the cell cycle is deregulated and G1 is frequently shortened, and whether it can be pharmacologically exploited for therapy, has not been tested. Here we show that the duration of G1 is a causal determinant of chromatin state in cancer cells and that extending G1 creates a therapeutic window for epigenetic drugs. Using acute myeloid leukemia (AML) as a model, we demonstrate that low-dose palbociclib, at concentrations well below those required for cytostatic arrest, extends G1 without halting proliferation. This modest prolongation reshapes the histone modification landscape: repressive marks (H3K9me2/3, H4K20me2/3) increase while acetylation decreases, and chromatin accessibility rises broadly in the euchromatic compartment. Naturally slow-cycling AML lines share this epigenetic signature regardless of their oncogenic driver mutations, and pharmacologically extending G1 in fast-cycling cells recapitulates it, establishing G1 length as a causal regulator of the cancer epigenome rather than a passive correlate. To identify epigenetic vulnerabilities created by G1 extension, we performed complementary drug and CRISPR-Cas9 screens in G1-extended AML cells. Both approaches converged on LSD1 (KDM1A): slow-cycling AML cells are intrinsically sensitive to LSD1 inhibition, while fast-cycling cells become sensitive when G1 is prolonged. The combination of low-dose palbociclib and LSD1 inhibition triggers differentiation and significantly prolongs survival in AML xenograft models. p21 (CDKN1A) emerges as the central molecular determinant of this response. In slow-cycling AML cells, p21 is highly expressed and its knockdown abolishes LSD1 inhibitor sensitivity. Structure-function analysis using p21 mutants separates the two known activities of p21: the CDK-inhibitory function (which extends G1) is required for sensitization, whereas the PCNA-binding function is dispensable. Three pharmacological routes converge on the same endpoint, CDK inhibition, G1 extension, and a differentiation-competent chromatin state: direct CDK4/6 inhibition by palbociclib, p21 overexpression, and p21 induction through HDAC or EZH1/2 inhibitors. Palbociclib bypasses the requirement for p21 entirely, confirming that G1 length itself, not p21 as a protein, is the critical variable. Mechanistically, the combination of G1 extension and LSD1 inhibition produces a qualitatively distinct chromatin state rather than an additive one. ATAC-seq reveals thousands of combination-exclusive accessible regions, enriched for footprints of myeloid differentiation transcription factors including SPI1/PU.1, IRF1, and STAT1/2. A double-lock principle governs this remodeling: palbociclib drives the removal of repressive marks (H3K9me3 and H3K27me3), while LSD1 inhibition installs active marks at the newly accessible regions. The ncBAF chromatin remodeling complex, identified in our CRISPR screen and validated by knockout of its essential subunits BRD9 and SMARCD1, is specifically required for this response. Loss of ncBAF abolishes the combination-induced chromatin remodeling and differentiation program but does not affect the initial G1 extension or retinoic acid-induced differentiation, indicating that ncBAF specifically couples cell-cycle modulation to chromatin remodeling rather than acting as a general differentiation factor. The principle generalizes beyond AML. In melanoma, breast cancer, and small-cell lung cancer (SCLC), sensitivity to LSD1 inhibition tracks with p21 expression and cycling speed. Primary melanoma samples stratified by p21 recapitulate the same pattern: p21-high, slow-cycling cells are sensitive; p21-low, fast-cycling cells are resistant but can be sensitized by palbociclib cotreatment. Cisplatin-induced drug-tolerant persister (DTP) cells, which emerge as a slow-cycling, chemo-resistant population and upregulate both p21 and LSD1, become vulnerable to LSD1 inhibition and are eradicated by the combination. In melanoma patient-derived xenograft (PDX) models, p21-high tumors respond to LSD1 inhibitor monotherapy, while p21-low tumors are sensitized by palbociclib cotreatment, with p21 knockdown abolishing the response. Together, these findings establish cell-cycle duration as a tunable regulator of the cancer epigenome and demonstrate that pharmacological G1 extension converts cytostatic CDK4/6 inhibition into an epigenetic sensitization strategy. Both fast-proliferating and slow-cycling tumor compartments (including drug-resistant persisters) can be targeted by matching the epigenomic state to the appropriate combination of cell-cycle modulators and epigenetic drugs. p21 emerges as a candidate biomarker for patient stratification. More broadly, our work repositions the cell cycle from a passive conduit for proliferation signals to an active, druggable regulator of chromatin fate, with implications that extend from cancer therapy to stem cell biology and regenerative medicine.

Circulating cell-free DNA (cfDNA) offers a minimally invasive lens into temporal tumor evolution. However, the accurate quantification of clonal composition from cfDNA remains challenging, particularly in low tumor fraction (TF) settings. Existing liquid biopsy deconvolution frameworks are frequently constrained by their reliance on bulk tissue references, simplified copy-number assumptions, and incomplete representations of clonal architecture, which collectively compromise sensitivity and bias evolutionary inferences. To address these limitations, we developed cfClone, a Bayesian framework that integrates single-cell whole-genome sequencing (scWGS) derived clonal structures with cfDNA whole-genome sequencing data to enable high-resolution, tissue-informed clonal tracking. Notably, while cfClone inherently leverages genomic instability, we demonstrate that cfClone achieves accurate TF estimates and circulating tumor DNA (ctDNA) detection even in malignancies with limited copy-number variant (CNV) burden. We validate cfClone in low and high CNV burden cases using simulated data derived from real patient data, establishing sensitive detection thresholds across a range of aneuploidy levels. By jointly modeling local copy-number alterations and allele-specific signals via Bayesian model selection and Markov chain Monte Carlo (MCMC) sampling, the algorithm yields uncertainty-aware estimates of clonal prevalence and TF. Applied to longitudinal clinical cohorts, cfClone reconstructs real-time evolutionary trajectories and uncovers clonal selection driving therapeutic resistance, including the de novo detection of emergent clonal populations. Github link: https://github.com/Roth-Lab/cfclone

Pancreatic ductal adenocarcinoma (PDAC) progression involves malignant cell state plasticity. Epigenetic changes underlie this plasticity, yet the PDAC cis-regulatory landscape remains understudied. To address this, we profiled 33 primary tumors and 7 metastases from 39 patients with single-cell ATAC-seq, paired with 10 single-cell RNA-seq profiles. We found that epigenetic GATA6+/KRT17+ co-accessibility identifies a classical-basal subtype-intermediate progenitor state (SIP) associated with better clinical outcomes. SIP cells display limited epigenetic reprogramming from premalignant epithelium and retain gastric-intestinal differentiation reminiscent of neoplastic precursors. Lineages without GATA6+/KRT17+ co-accessibility exhibit greater lineage and epithelial-mesenchymal plasticity. Classical PDACs that repress basal gene accessibility activate neural-like progenitor (NRP) and tuft lineage enhancers, whereas basal committed tumors display esophageal transdifferentiation. Compared to SIP, classical-NRP and basal committed tumors have poorer outcomes, and show distinct PD-1/PD-L1 immune proteomic phenotypes and prognostic myofibroblast epigenetic states, respectively. Our work reveals links between lineage reprogramming, EMT, and epigenetic progression in human PDAC.

Acute myeloid leukemia (AML) is sustained by leukemic stem cells (LSCs), which must be eradicated for durable remission yet remain therapy-resistant. Here, we present a scalable platform to identify compounds that eliminate LSCs by performing the first large high-throughput drug screen directly on human LSC-enriched cells. Using this system, we screened 11 142 compounds and identified 20 inhibitors selective for LSCs. For three candidates, BIO-acetoxime, SJB2-043 and UMxxxxx03, we confirmed anti-LSC activity across multiple primary patient samples and validated efficacy through xenotransplantation assays. Single-cell RNA sequencing uncovered convergent and distinct anti-LSC mechanisms of action, including USP1 inhibition, which suppressed stemness and cell-cycle programs while promoting metabolic and inflammatory stress responses. Collectively, the LSC-enriched fraction was eliminated through apoptosis or differentiation. Together, this work establishes the first platform to directly screen LSCs at scale and identifies novel vulnerabilities and therapeutics capable of eliminating the root of AML.

Acute myeloid leukemia (AML) cells exhibit pronounced metabolic plasticity, yet how amino acid supply is coordinated to sustain leukemic metabolism remains poorly understood. Here, we show that AML cells catabolize extracellular proteins as a major source of amino acids through lysosomal degradation of albumin. This proteocatabolic activity supports anabolic processes and mitochondrial energy production and establishes a regulated, high-throughput regime of primary nitrogen-containing metabolites (nitrogen regimen). Sustained proteocatabolism inevitably generates ammonia, and we find elevated ammonia concentrations in bone marrow plasma from newly diagnosed AML patients that decline with effective induction therapy. Using metabolomics, isotope tracing and targeted genetic and pharmacological manipulations, we identify glutamate-ammonia ligase (GS/GLUL) as a central enzyme that buffers proteocatabolism-derived ammonia by stabilizing intracellular nitrogen homeostasis. Loss of GS function limits sustainable nitrogen handling capacity, thereby impairing leukemic proliferation and delaying disease progression in vivo. Together, our findings define extracellular protein catabolism as a regulating nitrogen management strategy in AML and reveal GS as a capacity-defining vulnerability of proteocatabolic growth.

KIF18A inhibition selectively kills cancer cells by inducing chromosome alignment defects that activate a spindle assembly checkpoint (SAC)-dependent mitotic arrest. The mechanisms by which initially sensitive cancer cells acquire resistance to KIF18A inhibitors (KIF18Ai), the vulnerabilities of resistant cells, and the initial genetic determinants of KIF18Ai sensitivity remain poorly understood. Using orthogonal CRISPR-Cas9 screening and long-term drug adaptation approaches, we identify two convergent resistance mechanisms. Resistant cells either partially override the SAC, permitting mitotic exit despite chromosome misalignment, or adapt spindle microtubule dynamics to restore chromosome alignment in the absence of KIF18A. Both strategies sustain mitotic progression without inducing KIF18Ai dependence or increasing sensitivity to other mitotic perturbations. In contrast, reduced activity of the mitotic exit regulators PP2A or APC/C enhances KIF18Ai sensitivity. Accordingly, Mps1 inhibitor-driven APC/C mutations enhance responsiveness to KIF18A inhibition. In conclusion, resistance to KIF18A inhibition emerges rapidly, yet distinct genetic contexts create exploitable vulnerabilities to KIF18Ai.