Imageomics uses machine learning to accelerate our understanding of biological traits and human disease processes. Some of the earliest imageomics applications used deep learning to assess human diseases. For example, retinal fundus images were analyzed to diagnose diabetic retinopathy. The imaging modality optical coherence tomography (OCT) is widely used to diagnose and monitor the progression of retinopathy in patients and preclinical models. The standardized instrumentation and image format of OCT lends itself to imageomics, but generalizable, automated pipelines for segmentation and quantitation of large numbers of OCT images are still in early development. Here, we present the automated, end-to-end pipeline Crop-OCT that extracts features from thousands of OCT images, while preserving their location within the eye. We used the Crop-OCT pipeline on a diverse dataset, including 13 genetic models of retinopathy, with more than 20,000 OCT images, which allowed us to analyze nearly 6 million measured features. The pipeline was generalized on an independent dataset that was analyzed in a blinded manner. The pipeline enabled us to monitor ocular changes associated with aging and progression of diseases, such as retinitis pigmentosa, Leber congenital amaurosis, achromatopsia, Stargardt disease, diabetic retinopathy, and age-related macular degeneration. We also characterized heterogeneity across animals and identified regional and focal lesions. Our pipeline will unify feature extraction for preclinical models of retinal disease and serve as a foundation for future multimodal data integration for artificial intelligence applications based on imageomics.

Abstract Over fifty percent of patients diagnosed with neuroblastoma (NBL) will relapse with therapy resistant tumors. The 5-year survival rate of NBL relapse is less than 10%, leaving a desperate need to understand how these tumors recur. NBL tumors are heterogeneous and made up of two major cell populations derived from proliferating sympathoblasts: mesenchymal (MES) and adrenergic (ADRN). The central hypothesis is that the MES cell population evades standard of care treatments and re-establishes tumor recurrence in NBL. To better understand the dynamic of ADRN and MES cell states in tumor recurrence, a reporter was designed to visualize cell states. Using a small region of super-enhancers driving ADRN or MES cell states upstream of a minimal promoter and green fluorescent protein (GFP), these reporters show promise in well-established cell lines. Four reporters were created, two specific to ADRN cells and two specific to MES cells. Validation of these reporters using flow cytometry, western blot analysis, and qPCR showed that MES reporters were robust and specific, while ADRN were constitutively active in both cell sates. Additionally, these reporters were tested in patient-derived cell lines. Preliminary data shows that the MES reporters may also function in these lines. From these studies, we anticipate determining the cell population resistant to chemotherapeutics and identify the cell state that drives tumor recurrence. There is a drastic need for understanding the cellular makeup and features of relapsed NBL tumors. These studies are designed to determine how the tumors repopulate and may lead to a novel way to treat relapsed NBL tumors. Citation Format: Samantha Turk, Melody Allensworth, Justina McEvoy, Marybeth Lupo, Jackie Norrie, Michael A. Dyer. Novel reporter tool for visualization of neuroblastoma cell state [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 698.

Ewing sarcoma (EWS) is a highly aggressive pediatric malignancy characterized by elevated expression of SLFN11, which impairs DNA repair by binding to and functionally inhibiting DNA repair complexes, thereby enhancing susceptibility to genotoxic therapies. However, relapse remains a major clinical challenge and is often accompanied by the emergence of therapeutic resistance linked to reduced SLFN11 expression. We hypothesized that SLFN11-deficient tumors undergo adaptive metabolic reprogramming to overcome chemosensitivity. Here, we leverage transcriptomic and metabolomic profiling in patient-derived EWS models to demonstrate that SLFN11 loss drives downregulated mitochondrial glycerol-3-phosphate dehydrogenase (GPD2) expression, higher accumulation of glycerol-3-phosphate, fatty acid unsaturation, and enhanced glycerophospholipid (GPL) biosynthesis. Subsequently, targeting GPL biosynthesis (FSG67) restored DNA-damaging agent (SN-38) sensitivity in SLFN11-deficient EWS model, revealing a potential metabolic vulnerability to overcome chemoresistance. Furthermore, SLFN11 knockout tumors exhibited an elevated phosphocholine/glycerophosphocholine ratio, offering a potential non-invasive diagnostic biomarker.

Abstract Introduction: ABBV-706 is an antibody-drug conjugate (ADC) composed of a monoclonal antibody targeting the seizure-related homolog 6 (SEZ6) surface protein linked to topoisomerase 1 inhibitor payload. A Phase 1 trial of ABBV-706 is underway in adults with SEZ6 expressing relapsed solid tumors (NCT05599984). ABBV-706 became of interest as SEZ6 is highly expressed in many pediatric cancers. The goal was to evaluate ABBV-706 in models of childhood cancers with notable SEZ6 expression: neuroblastoma (NB), medulloblastoma (MB), retinoblastoma (RB), and alveolar rhabdomyosarcoma (ARMS) and compare the activity to vehicle and isotype non-target control ADC (IC-ADC). Methods: Pediatric xenograft models were screened for SEZ6 expression. Single agent in vivo studies were completed in NB, MB, RB, and ARMS models with varying SEZ6 expression. After engraftment, 2 total doses of ABBV-706 or IC-ADC were given by IP injection 3 weeks apart at 2 dose levels (DL1, DL2). Efficacy was assessed by median event free survival (KM med, in days) and objective response measure (Ped Blood Cancer 2007;49:928-940). ORM defines an objective response as partial, complete, or maintained complete response (PR, CR, and MCR) compared to stable disease (SD) or progressive disease, with or without growth delay (PD2 and PD1, respectively). Results: ABBV-706 showed efficacy in a broad range of SEZ6 expressing non-CNS (NB, RB, ARMS) models. At DL1, PR, CR, or MCR responses were seen in 8 of 11 models. At DL2, 9 of 11 models had PR, CR, or MCR responses. For most models, ABBV-706 was more active than IC-ADC. Survival advantage was observed in 1 of 2 MB models. Conclusion: We showed high anti-tumor activity for ABBV-706 in multiple SEZ6 expressing models and enhanced efficacy versus IC-ADC. Due to high expression of SEZ6 in several pediatric histologies, ABBV-706 has potential for clinical activity in patients with these cancers. Citation Format: Elizabeth A. Stewart, Michael A. Dyer, Yael P. Mosse, John M. Maris, Xiao-Nan Li, Stefan Atkinson, Chen He, David N. Groff, Alvin Farrel, Yuchen Du, Jinnan Chen, Steven B. Neuhauser, Timothy M. Stearns, Emily L. Jocoy, Jee Young Kwon, Jeffrey H. Chuang, Emily Faivre, Kelly Doyle, Kimberly E. Ellison, Joshua F. Hernandez, Joann P. Palma, Michael Barnes, Pooja Hingorani, Nandini Rudra-Ganguly, Beverly A. Teicher, Carol J. Bult, Malcolm A. Smith. Efficacy of ABBV-706, a SEZ6-targeted topoisomerase 1 inhibitor ADC: A report from the Pediatric Preclinical In Vivo Testing (PIVOT) Program [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 7815.

Ewing sarcoma (EWS) is an aggressive cancer in adolescents and young adults with frequent relapse rates and poor outcomes in recurrent or metastatic cases. Schlafen family member 11 ( SLFN11 ) gene is associated with the sensitivity to DNA-damaging agents (DDAs). The knockout of SLFN11 is associated with acquired chemoresistance in both cell lines and preclinical models. Here, we aimed to elucidate the metabolic underpinnings of SLFN11 -loss associated chemoresistance in patient derived cell lines of EWS. Our integrated transcriptomic and metabolomic analyses revealed downregulation of mitochondrial glycerol-3-phosphate dehydrogenase 2 ( GPD2 ) gene, which was accompanied by the upregulation of glycerophospholipid (GPL) biosynthesis pathway. Further, therapeutic targeting of lipid synthesis with the glycerol-3-phosphate acyltransferase 1 (GPAT1) inhibitor (FSG67) enhanced the efficacy of the DDA (SN-38) in SLFN11−/− cells. These findings indicate that SLFN11 loss-mediated chemoresistance can be targeted by blocking GPL biosynthesis in addition to DDA administration. ### Competing Interest Statement The authors have declared no competing interest. * EWS : ewing sarcoma SLFN11 : schlafen family member 11 DDA : DNA-damaging agent GPL : glycerophospholipid GPAT1 : glycerol-3-phosphate acyltransferase 1 GPD2 : glycerol-3-phosphate dehydrogenase 2 EWSR1 : ewing sarcoma breakpoint region 1 ETS : erythroblast transformation specific FLI1 : Friend Leukemia Integration 1 PARP : poly (ADP-ribose) polymerase G3P : glycerol-3-phosphate CCLE : Cancer Cell Line Encyclopedia DepMap : cancer dependency map CRISPR : clustered regularly interspaced short palindromic repeats RNAi : RNA interference FDR : false discovery rate TCGA : The Cancer Genome Atlas Program ESCLA : Ewing Sarcoma Cell Line Atlas RNA-seq : RNA sequencing PCA : principal component analysis PC1 : first principal component Log2FC : Log2 fold change WT : wild type SLFN11−/− : SLFN11 knock out Pearson R : Pearson correlation coefficient G3PS : glycerol-3-phosphate shuttle DHAP : dihydroxyacetone phosphate ETC : electron transport chain LC/MS : Liquid chromatography–mass spectrometry PE : phosphatidylethanolamine PC : phosphatidylcholine PI : phosphatidylinositol PG : phosphatidylglycerol PA : phosphatidic acid DAG : diacylglycerol AGPAT4 : 1-Acylglycerol-3-Phosphate O-Acyltransferase 4 MUFA : monounsaturated fatty acid 1H NMR : proton nuclear magnetic resonance PCh : phosphocholine GPC : glycerophosphocholine BRAID : Bivariate Response to Additive Interacting Doses GPAT1 : glycerol-3-phosphate acyltransferase 1 IAE : Index of Achievable Efficacy IDMA : inhibitory dose for the anchor drug IDMB : inhibitory dose for the partner drug EC50 : half maximal effective concentration RPA1 : replication protein A MCM3 : minichromosome maintenance complex component 3 CDC45 : cell division cycle 45 PCNA : proliferating cell nuclear antigen ATR : Ataxia Telangiectasia and Rad3-related CDT : chromatin licensing and DNA replication factor 1 DDB1 : damage-binding protein 1 CUL4 : cullin 4 NSCLC : non-small cell lung cancer ccRCC : clear cell renal cell carcinoma mTOR : mammalian target of rapamycin FASN : fatty acid synthase ACC : acetyl-CoA carboxylase SCD1 : stearoyl-CoA desaturase 1. American Lebanese Syrian Associated Charities, https://ror.org/03rx10x31

Retinoblastoma is a rare and aggressive pediatric tumor of the developing retina that originates in utero following biallelic inactivation of the tumor suppressor gene RB1. The early onset and biological complexity of this disease present unique challenges for researchers and clinicians, including difficulties in diagnosis and variability in treatment outcomes. Tumor heterogeneity in retinoblastoma arises from cellular, genetic, and developmental differences. Patients with germline RB1 mutations, and to a lesser extent those without, often develop multifocal tumors, though the origins and relationships between these lesions have remained unclear. Histopathological studies suggest that retinoblastoma tumors consist of two major cellular populations resembling either photoreceptors or retinal progenitor cells. However, molecular characterization of retinoblastoma tumors has historically been limited to cases of advanced disease requiring enucleation of the eye, restricting opportunities for comprehensive research on earlier stages of tumor development. To address this challenge, retinal organoids differentiated from stem cells were used to model early developmental stages of the retina and to study tumor initiation in a controlled environment. Patient-derived xenograft models were employed to investigate tumor progression, heterogeneity, and response to treatment in advanced disease. Additionally, single-cell RNA sequencing and cellular barcoding enabled detailed analysis of the developmental trajectory and differentiation status of tumor cell populations, revealing their roles in disease initiation, progression, and therapeutic outcomes. Our study definitively demonstrates the relationship between two the major cellular populations in retinoblastoma tumors: photoreceptor-like and progenitor-like cells. We further reveal that chemotherapy preferentially targets one of these populations, challenging traditional assumptions about treatment mechanisms and highlighting the influence of development and differentiation status on therapeutic outcomes. These findings underscore the critical importance of developmental heterogeneity in retinoblastoma biology and its influence on treatment response. By redefining the interplay between tumor composition and therapeutic outcomes, our work provides a foundation for developing more precise and effective treatment strategies for this devastating childhood cancer. Shannon R Sweeney, Madison Parks, Jackie Norrie, Asha Jacob Jannu, Cody Ramirez, Michael A Dyer. THROUGH THE LENS OF TUMOR HETEROGENEITY: UNRAVELING DEVELOPMENTAL DYNAMICS AND CHEMOTHERAPY RESISTANCE IN RETINOBLASTOMA [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Discovery and Innovation in Pediatric Cancer— From Biology to Breakthrough Therapies; 2025 Sep 25-28; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2025;85(18_Suppl_2):Abstract nr A012.

ABSTRACT:Sickle cell disease (SCD) is a common, life-threatening group of disorders caused by missense mutations in the β-globin gene (HBB). Mouse models have helped to elucidate the most common form of SCD (hemoglobin SS [HbSS]; homozygous p.Glu6Val) and develop new therapies. In contrast, a lack of animal models has restricted research on the second most common form of SCD (hemoglobin SC [HbSC]; p.Glu6Val/p.Glu6Lys). We used CRISPR genome engineering to generate HbSC alleles in the Townes mouse strain, which harbors human α- and β-globin genes in place of the mouse counterparts. Compared to Townes HbSS mice, HbSC mice exhibited signature pathologies that distinguish HbSC disease in humans.

The retina is a dynamic neural tissue that lines the posterior of the eye cup and transfers visual inputs from the world to our brain for processing. However, exposure to stress, injury, and disease can disrupt this important function. Here, we present a protocol for the dissection and culture of retina in various physiologically relevant stress conditions. Furthermore, we describe how to assess the retina for stress and cell-type-specific responses. For complete details on the use and execution of this protocol, please refer to Norrie et al.1.

Retinoic acid (RA) is a standard-of-care neuroblastoma drug thought to be effective by inducing differentiation. Curiously, RA has little effect on primary human tumors during upfront treatment but can eliminate neuroblastoma cells from the bone marrow during post-chemo maintenance therapy—a discrepancy that has never been explained. To investigate this, we treat a large cohort of neuroblastoma cell lines with RA and observe that the most RA-sensitive cells predominantly undergo apoptosis or senescence, rather than differentiation. We conduct genome-wide CRISPR knockout screens under RA treatment, which identify bone morphogenic protein (BMP) signaling as controlling the apoptosis/senescence vs differentiation cell fate decision and determining RA’s overall potency. We then discover that BMP signaling activity is markedly higher in neuroblastoma patient samples at bone marrow metastatic sites, providing a plausible explanation for RA’s ability to clear neuroblastoma cells specifically from the bone marrow, by seemingly mimicking interactions between BMP and RA during normal development.

Alveolar rhabdomyosarcoma (ARMS) patients harboring paired-box fusion proteins (PAX3/7-FOXO1) exhibit a greater incidence of tumor relapse, metastasis, and poor survival outcome, thereby underscoring the urgent need to develop effective therapies to treat this subtype of childhood cancer. To uncover mechanisms that contribute to tumor initiation, we develop a muscle progenitor model and use epigenomic approaches to unravel genome rewiring events mediated by PAX3/7 fusion proteins. Among the key targets of PAX3/7 fusion proteins, we identify a cohort of oncogenes, fibroblast growth factor (FGF) receptors, tRNA-modifying enzymes, and genes essential for mitochondrial metabolism and protein translation, which we successfully targeted in preclinical trials. We identify leucine usage as a key factor driving the growth of aggressive PAX-fusion tumors, as limiting its bioavailability impaired oxidative phosphorylation and mitochondrial metabolism, delaying tumor progression and improving survival in vivo. Our data provide a compelling list of actionable targets and suggest promising new strategies to treat this tumor.

TERT is a bona fide MYCN target. A, Alignments of DNA methylation from WGBS and MYCN-ChIP-seq in three MYCN-amplified neuroblastoma O-PDX models and cell line (SKNBE2) showing that MYCN binding overlaps with the hypomethylated region in the TERT promoter. The core promoter is shaded in red and THOR is shaded in yellow. B, Western blot analysis for MYCN and GAPDH (loading control) in different indicated doxycycline-inducible engineered cell lines to ectopically express MYCN in non–MYCN-amplified cells (left) or to express three different MYCN-shRNAs (indicated as A, B, C) in MYCN-amplified neuroblastoma cells (right). Relative MYCN band intensities in cells expressing MYCN-shRNA are indicated on top of the blots. C, A schematic representation of Luciferase reporter construct. The long construct contains the luciferase gene, and the TERT promoter core and THOR sequences, while the short construct does not contain the THOR promoter sequence. D, Fold change of Luciferase luminescence for the short construct (top) and the long construct (bottom) compared with control vehicle-treated controls. E, Fold change of TERT expression in the same cells measured by qRT-PCR. *, P < 0.05, error bars represent SD.

SUMMARY Neuroblastoma is a pediatric cancer arising from the developing sympathoadrenal lineage with complex inter- and intra-tumoral heterogeneity. To chart this complexity, we generated a comprehensive cell atlas of 55 neuroblastoma patient tumors, collected from two pediatric cancer institutions, spanning a range of clinical, genetic, and histologic features. Our atlas combines single-cell/nucleus RNA-seq (sc/scRNA-seq), bulk RNA-seq, whole exome sequencing, DNA methylation profiling, spatial transcriptomics, and two spatial proteomic methods. Sc/snRNA-seq revealed three malignant cell states with features of sympathoadrenal lineage development. All of the neuroblastomas had malignant cells that resembled sympathoblasts and the more differentiated adrenergic cells. A subset of tumors had malignant cells in a mesenchymal cell state with molecular features of Schwann cell precursors. DNA methylation profiles defined four groupings of patients, which differ in the degree of malignant cell heterogeneity and clinical outcomes. Using spatial proteomics, we found that neuroblastomas are spatially compartmentalized, with malignant tumor cells sequestered away from immune cells. Finally, we identify spatially restricted signaling patterns in immune cells from spatial transcriptomics. To facilitate the visualization and analysis of our atlas as a resource for further research in neuroblastoma, single cell, and spatial-omics, all data are shared through the Human Tumor Atlas Network Data Commons at www.humantumoratlas.org .

Abstract Retinoic acid (RA) is a standard-of-care neuroblastoma drug used during post-chemo consolidation therapy. Based on clinical trials from the 90s, RA benefits 10-15% of patients. It is widely believed the anti-cancer activity of RA is due to retinoid-induced differentiation of cancer cells, conclusions largely attributable to observations in cell culture. However, given RA is typically used in the minimal residual disease setting, this mechanism has never been definitively proved in patients. To better understand RA’s activity, we deployed several new technologies. First, we conducted genome-wide CRISPR modifier screens in RA-treated hyper-sensitive neuroblastoma cell lines. Surprisingly, we found that, in these cells, RA primarily decreased cell viability via apoptosis or senescence, rather than differentiation—activities in which the CRISPR screens strongly implicated bone morphogenetic protein (BMP) signaling. Using ChIP-seq and RNA-seq we showed these behaviors were mediated by the coordinated gene regulatory actions of RARA and BMP-family SMAD transcription factors. Notably, interactions between BMP signaling and RA are well established in developmental biology, where BMP signaling can tip cell fate decisions between differentiation, apoptosis, and senescence upon exposure to naturally occurring RA, behaviors that can seemingly be maintained in neuroblastoma cells. Next, we assessed the correlations between RA IC50 and the expression of all (∼20,000) genes in a panel of 19 cell lines. Remarkably, SMAD9, a critical downstream transcription factor of the BMP pathway, was the #1 most correlated gene with RA IC50 (R = -0.92, P = 4.2 × 10-6), suggesting a highly generalizable relationship between BMP signaling and RA response. By modulating SMAD9, or other components of BMP signaling, we could promote apoptosis/senescence and sensitize cells to RA. We then performed large-scale drug combination screens of RA and the drug FK506, which can amplify BMP signaling activity. RA exhibited a synergistic effect with FK506 in all 10 neuroblastoma cell lines we tested, very strikingly in some cases, suggesting it could be possible to pharmacologically amplify the activity of RA in patients. Finally, using published single-cell RNA-seq data from neuroblastoma patient samples, we found BMP signaling activity is relatively low in primary tumors, but much higher in disseminated metastatic neuroblastoma cells in the bone marrow. We confirmed this trend using immunofluorescence staining in 6 paired primary patient samples. This site-specific variability in BMP signaling activity provides the first reasonable explanation for RA’s curious clinical activity, whereby it has little effect on bulky established tumors, but has been shown to clear disseminated metastatic cells from the bone marrow during consolidation therapy. Overall, our study revealed that BMP signaling controls neuroblastoma cell fate and sensitivity to RA and that this observation is consistent with the unique clinical behaviors of this drug. Citation Format: Min Pan, Yinwen Zhang, William C. Wright, Hyeong-Min Lee, Richard H. Chapple, Xueying Liu, Jonathan Low, Duane Currier, Allister J. Loughran, Michael A. Dyer, Shondra M. Pruett, Burgess Freeman III, Taosheng Chen, Brian J. Abraham, Elizabeth Stewart, John Easton, Paul Geeleher. BMP signaling determines neuroblastoma sensitivity to retinoic acid by directing cell fate [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Advances in Pediatric Cancer Research; 2024 Sep 5-8; Toronto, Ontario, Canada. Philadelphia (PA): AACR; Cancer Res 2024;84(17 Suppl):Abstract nr B005.

Programs for preclinical testing of targeted cancer agents in murine models of childhood cancers have been supported by the National Cancer Institute (NCI) since 2004. These programs were established to work collaboratively with industry partners to address the paucity of targeted agents for pediatric cancers compared with the large number of agents developed and approved for malignancies primarily affecting adults. The distinctive biology of pediatric cancers and the relatively small numbers of pediatric cancer patients are major challenges for pediatric oncology drug development. These factors are exacerbated by the division of cancers into multiple subtypes that are further sub-classified by their genomic properties. The imbalance between the large number of candidate agents and small patient populations requires careful prioritization of agents developed for adult cancers for clinical evaluation in children with cancer. The NCI-supported preclinical pediatric programs have published positive and negative results of efficacy testing for over 100 agents to aid the pediatric research community in identifying the most promising candidates to move forward for clinical testing in pediatric oncology. Here, we review and summarize lessons learned from two decades of experience with the design and execution of preclinical trials of antineoplastic agents in murine models of childhood cancers.

Alveolar rhabdomyosarcoma (ARMS) patients harboring PAX3-FOXO1 and PAX7-FOXO1 fusion proteins exhibit a greater incidence of tumor relapse, metastasis, and poor survival outcome, thereby underscoring the urgent need to develop effective therapies to treat this subtype of childhood cancer. To uncover mechanisms that contribute to tumor initiation, we developed a novel muscle progenitor model and used epigenomic approaches to unravel genome re-wiring events mediated by PAX3/7 fusion proteins. Importantly, these regulatory mechanisms are conserved across established ARMS cell lines, primary tumors, and orthotopic-patient derived xenografts. Among the key targets of PAX3- and PAX7- fusion proteins, we identified a cohort of oncogenes, FGF receptors, and genes essential for mitochondrial metabolism and protein translation, which we successfully targeted in preclinical trials. Our data suggest an explanation for the relative paucity of recurring mutations in this tumor, provide a compelling list of actionable targets, and suggest promising new strategies to treat this tumor.

Background Neuroblastoma is a common pediatric cancer, where preclinical studies suggest that a mesenchymal-like gene expression program contributes to chemotherapy resistance. However, clinical outcomes remain poor, implying we need a better understanding of the relationship between patient tumor heterogeneity and preclinical models.Results Here, we generate single-cell RNA-seq maps of neuroblastoma cell lines, patient-derived xenograft models (PDX), and a genetically engineered mouse model (GEMM). We develop an unsupervised machine learning approach ("automatic consensus nonnegative matrix factorization" (acNMF)) to compare the gene expression programs found in preclinical models to a large cohort of patient tumors. We confirm a weakly expressed, mesenchymal-like program in otherwise adrenergic cancer cells in some pre-treated high-risk patient tumors, but this appears distinct from the presumptive drug-resistance mesenchymal programs evident in cell lines. Surprisingly, however, this weak-mesenchymal-like program is maintained in PDX and could be chemotherapy-induced in our GEMM after only 24 h, suggesting an uncharacterized therapy-escape mechanism.Conclusions Collectively, our findings improve the understanding of how neuroblastoma patient tumor heterogeneity is reflected in preclinical models, provides a comprehensive integrated resource, and a generalizable set of computational methodologies for the joint analysis of clinical and pre-clinical single-cell RNA-seq datasets.

Abstract While retinoic acid (RA) has been successfully used for leukemia treatment for decades, the attempt to treat solid tumors with RA remains challenging, with less than 10% of neuroblastoma (NB) patients achieving complete remission when treated with RA alone. It has been believed the anti-cancer activity of retinoids is due to retinoid-induced terminal differentiation, however, what determines cell response or whether differentiation is the main effect of RA has never been fully understood. To better understand RA’s activity, we conducted genome-wide CRISPR modifier screens in RA-treated hyper-sensitive NB cell lines. Surprisingly, we found that, in these cells, RA primarily decreased cell viability via apoptosis or senescence, rather than differentiation—activities in which the CRISPR screens strongly implicated bone morphogenetic protein (BMP) signaling. BMP activation promoted apoptosis/senescence and sensitized cells to RA. Conversely, BMP inhibitors and SMAD9 (a critical transcription factor of the BMP pathway) knockout enhanced RA’s ability to induce differentiation but reduced cell sensitivity to RA. Our ChIP-seq and RNA-seq data showed these behaviors were mediated by the coordinated gene regulatory actions of RARA and BMP-family SMAD transcription factors. Furthermore, in a panel of 19 cell lines we screened with RA, we assessed the correlations between RA IC50 and the expression of all (~20,000) genes (publicly available data from the GDSC and Depmap). Remarkably, SMAD9 was the number 1 and number 10 most correlated gene, respectively, with RA IC50 (ranked by Pearson correlation coefficient; GDSC R = -0.92, P = 4.2 × 10−6; Depmap R = -0.81, P = 4.8 × 10−4). We also performed comprehensive large-scale drug screens in these cell lines with combinations of RA and a BMP activator FK506. RA exhibited a synergistic effect with FK506 in all the NB cell lines, very strikingly in some cell lines. All these data suggest BMP signaling is generally required for RA sensitivity and BMP activators are promising candidates in combination with RA to treat NB. Using published bulk and single cell RNA-seq data from NB patient samples, we found BMP signaling activity is relatively low in primary tumors, but well maintained in disseminated NB cells derived from bone marrow. This explains why RA is clinically used as a maintenance therapy and can only successfully treat the minimal residual disease and suggests that tumor microenvironment is a critical factor in determining cell response to RA. Overall, our study revealed that BMP controls NB cell fate and sensitivity in RA treatment. Notably, interactions between BMP signaling and RA are well established in developmental biology, where BMP signaling can tip cell fate decisions between differentiation, apoptosis, and senescence upon exposure to endogenous RA. Our data suggest that this developmental process can be maintained in NB cells, unveiling a novel mode of anti-neoplastic action. Citation Format: Min Pan, Yinwen Zhang, William C. Wright, Hyeong-Min Lee, Richard H. Chapple, Xueying Liu, Jonathan Low, Duane Currier, Allister J. Loughran, Dyer A. Dyer, Shondra M. Pruett, Burgess Freeman, Taosheng Chen, Brian J. Abraham, Elizabeth Stewart, John Easton, Paul Geeleher. BMP signaling determines neuroblastoma cell fate and sensitivity to retinoic acid [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 151.

Rhabdomyosarcoma (RMS) is a pediatric tumor that resembles undifferentiated muscle cells; yet the extent to which cell state heterogeneity is shared with human development has not been described. Using single-cell/nucleus RNA sequencing from patient tumors, patient-derived xenografts, primary in vitro cultures, and cell lines, we identify four dominant muscle-lineage cell states: progenitor, proliferative, differentiated, and ground cells. We stratify these RMS cells/nuclei along the continuum of human muscle development and show that they share expression patterns with fetal/embryonal myogenic precursors rather than postnatal satellite cells. Fusion-negative RMS (FN-RMS) have a discrete stem cell hierarchy that recapitulates fetal muscle development and contain therapy-resistant FN-RMS progenitors that share transcriptomic similarity with bipotent skeletal mesenchymal cells. Fusion-positive RMS have tumor-acquired cells states, including a neuronal cell state, that are not found in myogenic development. This work identifies previously underappreciated cell state heterogeneity including unique treatment-resistant and tumor-acquired cell states that differ across RMS subtypes.