Osteosarcoma (OS) is the most common human primary bone cancer, primarily affecting children and young adults. While the survival rate of patients diagnosed with localized OS is ∼65%, this decreases to ∼20% for patients with metastatic disease, and recurrent disease remains largely incurable. Therefore, identifying new therapeutic strategies is urgently needed for metastatic and refractory OS. This study analyzes the surfaceomes and global proteomes of 22 unique OS patient-derived xenografts (PDXs) using mass spectrometry to identify surface proteins for potential immunotherapeutic targeting. Both methods identify known OS-associated surface candidates including LRRC15, MMP14, MRC2, and CADM1, and several poorly characterized targets, including ROR2 and TMEM119. Both ROR2 and TMEM119 display robust expression in OS tissues but only limited or no expression in normal pediatric tissues, and loss of both targets reduces the migration of OS cells. These data provide a resource of surface proteins as potential immunotherapeutic targets in OS.

Abstract Background: MYCN amplification (MNA) is a defining biomarker of highrisk neuroblastoma with a poor clinical outcome. While tumorbased wholegenome sequencing (WGS) coupled with copynumber analysis (CNA) reliably identifies MNA, its detection in blood has remained uncertain due to low circulating tumor cell (CTC) burden. There is increasing interest in exploiting structural variants (SVs), which exhibit extremely low sequencing error rates, to enable ultra-sensitive tumor detection in liquid biopsies. Methods: We analyzed MNA neuroblastoma cases across three institutional cohorts with available paired tumor and peripheral blood (PB) WGS data. Mathematical modeling was performed to model limit of detection in using CNA versus SVbased approaches under varying tumor purities and sequencing depths. CNA detection was performed using CNVkit, ichorCNA, and DELLY, whereas tumorspecific SVs were genotyped in matching PB with our recently published error suppression method SVInDelGenotyper, and their error profiles are investigated using gnomAD and a healthy control cohort SJLIFE. Buccal samples and blood samples at remission were evaluated as alternative germline references. Results: CNAbased detection identified MNA in 32% of PB samples (9/28), with methodspecific sensitivities ranging from 4-25%. Integration of tumorinformed SV genotyping uncovered additional MYCN-linked breakpoints in CNAnegative PB samples, increasing overall detection sensitivity to 64% (18/28). SVs remained detectable even when PB samples exhibited no visible copynumber elevation (diploidlevel coverage), revealing tumor fractions as low as ∼0.01%. Deeper WGS coverage and SV burden correlated with improved SV detection across cohorts. Our findings highlight the risk of missing detection of key mutations for childhood neuroblastoma by standard tumor vs normal comparison when diagnostic PB is used as normal control. Buccal samples and remissionstage blood showed no tumorspecific SVs, supporting their role as reliable germline controls. Conclusions: Standarddepth WGS (30-100×), when combined with SVbased liquidbiopsy analysis, enables sensitive detection of MNA neuroblastoma from blood, outperforming CNAonly approaches especially in lowpurity samples. These findings establish tumorspecific SV genotyping as a powerful noninvasive strategy for identifying highrisk MNA neuroblastoma. Our data further suggests the future study design of childhood neuroblastoma liquid biopsy to employ buccal samples as normal controls. Citation Format: Pandurang Kolekar, Rebecca S. Kaufman, Hanxia Li, Yanling Liu, Yuan Feng, Bo Wang, Xi Wang, Li Fan, Lu Wang, Jinghui Zhang, John M. Maris, Sharon J. Diskin, Xiaotu Ma. Sensitive detection of MYCN amplified neuroblastoma in blood samples with structural variants using standard whole genome sequencing [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(8_Suppl):Abstract nr LB016.

MYCN binding to GP2 and ODC1 and enhancer window ATAC.

Abstract Developing effective chimeric antigen receptor (CAR) T cell therapies for pediatric solid tumors requires discovery of highly tumor-selective cell surface molecules. Using a multimodal immunotherapeutic target discovery platform, we identified a receptor composed of the nicotinic acetylcholine receptor subunits α3 and β4 as a new previously unrecognized immunotherapeutic target in neuroblastoma. CHRNA3 (α3) and CHRNB4 (β4) are robustly differentially overexpressed in neuroblastoma compared with normal tissues, driven by a super-enhancer upstream of the CHRNA3/B4 locus on chromosome 15q25 that is occupied by neuroblastoma core regulatory transcription factors and physically engages both promoters. Neuroblastoma cells exhibit abundant α3 and β4 protein, which bind and assemble into a stabilized α3β4 complex displayed at high levels on the neuroblastoma cell surface. Functionally, genetic depletion of α3 also reduced β4 levels, induced mesenchymal cell state-associated marker expression, and increased proliferation and invasion in neuroblastoma cell line models. To therapeutically exploit this tumor-restricted α3β4 receptor expression, we engineered Dual Interacting Subunit COmplex (DISCO)-specific single chain variable fragments (scFvs) designed to recognize an epitope requiring simultaneous engagement of both α3 and β4 subunits of the α3β4 cell surface receptor. DISCO CAR T cells generated from these scFvs bound selectively to and were potently activated by the α3β4 receptor, but not either subunit individually. α3β4-targeted DISCO CAR T cells demonstrated potent and selective cytotoxicity across diverse neuroblastoma cell lines and three patient-derived xenograft (PDXs) models, significantly extending the survival of mice harboring PDXs with a range of α3 and β4 expression (P < 0.05). Beyond neuroblastoma, CHRNA3 and CHRNB4 are also highly expressed in small cell lung cancer (SCLC) and retinoblastoma, among several other tumors. The identical super-enhancer is associated with high levels of CHRNA3 and CHRNB4 in SCLC, suggesting similar mechanisms of tumor-driven overexpression. In an intraocular retinoblastoma xenograft model, a single intravitreal dose of α3β4-redirected DISCO CAR T cells achieved marked tumor control and significantly extended ocular survival (P = 0.0015), collectively showing disease-relevance and targetability beyond neuroblastoma. Together, these findings broaden the repertoire of actionable CAR T cell targets in pediatric and adult solid tumors by validating the α3β4 receptor, and more broadly the nicotinic acetylcholine receptor family, as bona fide immunotherapeutic candidates. Moreover, these studies demonstrate that CARs can be designed to recognize disease-relevant subunit configurations within multidomain surface receptors. Citation Format: Patrick M. Schürch, Vincent P. Zecchino, Anna M. Giudice, Rebecca S. Kaufman, Evan Cresswell-Clay, Guillem Pascual-Pasto, Sydney L. Roth, Brendan C. McIntyre, Rawan Shraim, Amber K. Hamilton, Kush Parikh, Karina L. Conkrite, Khanh B. Trang, Grant P. Grothusen, David Groff, Pamela Mishra, Simona Lombardi, Tyler Skinner, Andrew D. Wells, Struan F. Grant, Daniel Martinez, Wei Li, Sharon J. Diskin, Kristopher R. Bosse. Selective targeting of the α3β4 nicotinic acetylcholine receptor by DISCO (dual interacting subunit complex) CAR T cells [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 1332.

Despite intensive, multimodal therapy, only half of children diagnosed with high-risk neuroblastoma will survive 5 years, and survivors harbor significant short- and long-term treatment-related comorbidities. Although mAb therapy targeting GD2 has improved outcomes, GD2-directed immunotherapy remains one of the only FDA-approved immunotherapies for pediatric cancer, and therapy is toxic due to GD2 expression on pain fibers. Thus, there is a critical need to uncover new immunotherapy targets in neuroblastoma. Activated leukocyte cell adhesion molecule (ALCAM) is a cell adhesion molecule that promotes tumor growth in a variety of cancers and is highly expressed in neuroblastoma. We generated three inducible CRISPR inhibition cell lines to deplete ALCAM and elucidate its role in neuroblastoma. Depletion of ALCAM reduced cell growth, reduced Ki-67 staining, and increased cleaved PARP. To determine the mechanism of ALCAM overexpression, we used chromatin immunoprecipitation sequencing to show MYCN oncoprotein binding at the ALCAM promoter. We generated luciferase reporters from the ALCAM promoter and a putative upstream (10 kb) enhancer, which we defined using Promoter-based Capture-C. Treatment with the MYC(N)/MAX dimerization inhibitor MYCi975 reduced ALCAM expression by immunoblotting and luciferase signal from the ALCAM promoter. We validated the activity of the upstream enhancer and uncovered an AP-1-binding motif that is critical for enhancer activity. Finally, as ALCAM is expressed in several normal tissues, we investigated an ALCAM-targeted conditionally activated antibody-drug conjugate, CX-2009 (praluzatamab ravtansine), which delayed tumor growth in two of three patient-derived xenograft models. Together, these findings credential ALCAM as an immunotherapeutic target in neuroblastoma.

Abstract Background: We recently defined the NB/EWS surfaceomes using integrative proteogenomics to prioritize proteins as candidate immunotherapeutic targets. GFRA2 was a top ranked candidate for both NB and EWS (Clin Cancer Res 2023, Cancer Cell 2024). Aims: (1) Validate and assess mechanism of GFRA2 overexpression; (2) Identify selective antibody binders to GFRA2; (3) Engineer antibody drug conjugates (ADCs) and test for internalization and potency. Methods: ChIP- and RNA-sequencing of NB and RNA-sequencing following siRNA depletion of EWSR1::FLI1 were used to evaluate mechanisms of GFRA2 overexpression. GFRA2 abundance and cellular localization was evaluated by flow cytometry and immunofluorescence. Phage display was performed with recombinant GFRA2 extracellular domain protein as the bait and the highly conserved GFRA1 and GFRA3 recombinant proteins as counters. In parallel, we humanized the anti-GFRA2 murine HSAN antibody. While we plan on reporting on several linker-payload combinations, here we focus on the initial humanized HSAN antibody conjugated to pyrrolobenzodiazepine (PBD) via a cleavable linker, which was tested for internalization using live cell imaging and cytotoxicity across a panel of human NB and EWS preclinical models. Results: We identified a proximal super enhancer (Percentile: 97.4%-99.7%) or enhancer (94.4%-97.7%) in all 10 NB cell lines profiled. Depletion of EWSR1::FLI1 resulted in significantly decreased GFRA2 mRNA expression across 6 EWS cell lines (P<0.0001), and this was validated via immunoblotting. GFRA2 protein was expressed uniformly and to variable degrees on the cell surface of 10 NB cell lines, 8 NB patient-derived xenografts, and 6 EWS cell lines by flow cytometry and/or immunofluorescence. Patient tumor RNA-Seq showed 92.8% of NB (N=126; median=47.65) and 74% of EWS (N=85, median=11.47) had high expression defined as a GFRA2 TPM > 5. Phage display panning of fully human antibody fragments yielded three Fabs and one VH single domain antibody specific to GFRA2 with moderate binding affinity (50-500 nM). Humanized HSAN (co-hu-HSAN) showed high binding affinity specific to GFRA2 (1-10 nM). We then showed robust internalization of co-hu-HSAN across 6 NB cell lines. A co-hu-HSAN-PBD ADC showed potent and specific cytotoxicity across GFRA2+ cell lines (NB=2; median IC50=1.69;2.92pM, EWS=2; median IC50=9.81;12.43pM) with no cytotoxicity in GFRA2 null cell lines (N=2; median IC50=NA). Conclusions: GFRA2 is a lineage restricted oncoprotein abundantly expressed in both NB/EWS, as well as other human solid cancers. We show initial proof-of-concept for potent and specific cytotoxicity in both NB/EWS with the co-hu-HSAN-PBD ADC. We will report on efficacy testing of the co-hu-HSAN-PBD ADC across NB/EWS xenograft models. Work to affinity enhance the fully human binders, and progress with additional linker/payload combinations are ongoing and will be reported. Citation Format: Amber K. Hamilton, Seungmin Shin, Raphael D. Lopez, Nicholas Hartnett, Alexander B. Radaoui, Maggie Hines, Maria Evancho, Rebecca S. Kaufman, Khushbu Patel, Karina L. Conkrite, Dan Martinez, Brian Mooney, Michelle E. Keyel, Elissa Levine, Alberto D. Guerra, Jarrett Lindsay, Yael P. Mosse, Jennifer Pogoriler, Gregg Morin, Poul H. Sorensen, Patrick J. Grohar, Benjamin A. Garcia, C. Patrick Reynolds, Wei Li, Sharon J. Diskin, John M. Maris. Development of a GFRA2-targeting antibody drug conjugate for neuroblastoma (NB), and Ewing sarcoma (EWS) [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 7805.

Background:Pediatric brain tumors are the leading cause of disease-related mortality in children, yet many aggressive tumors lack effective therapies. RNA splicing is a hallmark of cancer, but it has not yet been systematically studied in pediatric brain tumors. Methods:We analyzed 729 pediatric brain tumors spanning histologies and molecular subtypes to quantify differential tumor splicing. We developed the Splicing Burden Index (SBI) to enable cross-sample comparisons and performed hierarchical clustering of highly variable splice events to define splicing-informed tumor groups. These were integrated with clinical outcomes, pathway activity, and proteogenomic data. Recurrent splice events were prioritized for predicted functional impact, and in vitro perturbation studies were performed targeting the splicing kinase CDC-like kinase 1 (CLK1). Results:SBI revealed substantial interhistology and intrahistology heterogeneity. Clusters were enriched for histologies and molecular subtypes, several of which were independently associated with survival beyond histology and clinical covariates. Spliceosome pathway activity varied across clusters and was associated with worse survival, yet was not correlated with SBI, indicating distinct dimensions of splicing dysregulation. Functional prioritization identified a recurrent in CLK1 exon 4, required for canonical kinase activity. CLK1 exon 4 inclusion followed an oncofetal pattern and showed context-dependent associations with outcome distinct from total CLK1 expression. Pharmacologic inhibition and exon 4-specific perturbation of CLK1 reduced tumor cell viability and disrupted cancer-relevant splicing and transcriptional programs. Conclusions:This study systematically characterizes splicing in pediatric brain tumors, identifies splicing-informed subgroups, and prioritizes CLK1 exon 4 as an oncofetal tumor-specific event, motivating further preclinical exploration.

Supplemental Tables including Representativeness of Study Participants

ALCAM expression in cell lines, developing adrenal gland and patient tumors.

Figure S5: Characteristics of lncMod analysis and results

Table S1: TARGET clinical sample and RNA-sequencing characteristics, Table S2: Genomic loci for lncRNA and protein coding genes in this study, Table S3: Number and types of genes expressed per cancer

Table S13: lncRNAs associated with CRC of NBL, Table S14: Differentially expressed lncRNAs between major subtypes in NBL

Aberrant pre-mRNA splicing is a hallmark of many cancers, yet the underlying genetic mechanisms driving these alterations remain incompletely understood. While splice-altering variants can explain a subset of events, many tumor-specific splicing changes arise in the absence of nearby sequence variants. This study aims to systematically dissect the contribution of proximal splice-altering variants to tumor-specific splicing in pediatric brain tumors and to define the residual cohort likely driven by regulatory mechanisms such as mutations affecting splicing factors or the spliceosome machinery. We hypothesized that some fraction of tumor-specific splice events in pediatric brain tumors are not associated with nearby genetic variants, and that these may highlight broader disruptions to splicing regulation. We applied replicate Multivariate Analysis of Transcript Splicing (rMATS-turbo) to stranded total RNA-seq data (N=635) from the Pediatric Brain Tumor Atlas (PBTA). Tumor-specific single exon (SE) splicing events were defined as a change in percent spliced in (| ΔPSI |) > 0.3 between tumors and healthy brain samples sourced from GTEx (<40 years, n=2,642) and pediatric normals (n=7). Tumor-normal matched whole genome sequencing data were used to identify somatic and rare (<0.1% Allele Frequency in gnomad) germline SNVs/Indels proximal (<250nt) to rMATS-defined splice junctions. Across the cohort (N=635), we identified 107,827 tumor-specific SE alternative splicing events. Only 44 tumors harbored a variant proximal to the affected splice site (11 germline, 39 somatic), leaving most events without a clear cause. We are now investigating this variant-negative subgroup as a preliminary cohort. Current efforts focus on excluding structural drivers before turning to potential regulatory mechanisms, including the disruption of spliceosome components. Preliminary findings suggest some tumors show altered expression or mutations in spliceosome components, though further validation is needed. This dual-ended analysis of tumor-specific splicing events reveals that while a subset can be attributed to local genetic variation, the majority likely arise from broader disruptions in splicing regulation. These findings underscore the importance of integrating RNA- and DNA-based analyses to fully capture the molecular basis of splicing dysregulation in cancer. Moreover, the variant-negative cohort may serve as a rich resource to uncover novel regulatory mechanisms and potential therapeutic targets in splicing-driven pediatric brain tumors. Patricia J Sullivan, Ryan J Corbett, Ammar S Naqvi, Alexander Sickler, Rebecca Kaufman, Bo Zhang, Chuwei Zhong, Sharon J Diskin, Jo Lynne Rokita. Uncovering genetic and regulatory drivers of tumor-specific splicing in pediatric brain tumors [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 B031.

Figure S3: Regions of somatic copy number aberration across cancers and genes dysregulated due to copy number

PURPOSE:DROSHA, DGCR8, and DICER1 regulate miRNA biogenesis and are commonly mutated in cancer. Although DGCR8 and DICER1 germline pathogenic variants (GPV) cause autosomal dominant tumor predisposition, no association between DROSHA GPVs and clinical phenotypes has been reported. EXPERIMENTAL DESIGN:After obtaining informed consent, sequencing was performed on germline and tumor samples from all patients. The occurrence of germline DROSHA GPVs was investigated in large pediatric and adult cancer datasets. The population prevalence of DROSHA GPVs was investigated in the UK Biobank and Geisinger DiscovEHR cohorts. RESULTS:We describe nine children from eight families with heterozygous DROSHA GPVs and a diagnosis of pineoblastoma (n = 8) or Wilms tumor (n = 1). A somatic second hit in DROSHA was detected in all eight tumors analyzed. All pineoblastoma tumors analyzed were classified as miRNA processing-altered 1 subtype. We estimate the population prevalence of germline DROSHA loss-of-function variants to be 1:3,875 to 1:4,843 but find no evidence for increased adult cancer risk. CONCLUSIONS:This is the first report of DROSHA-related tumor predisposition. As pineoblastoma and Wilms tumor are also associated with DICER1 GPVs, our results suggest that the tissues of origin for these tumors are uniquely tolerant of general miRNA loss. The miRNA processing-altered 1 pineoblastoma subtype is associated with older age of diagnosis and better outcomes than other subtypes, suggesting DROSHA GPV status may have important clinical and prognostic significance. We suggest that genetic testing for DROSHA GPVs be considered for patients with pineoblastoma, Wilms tumor, or other DICER1-/DGCR8-related conditions and propose surveillance recommendations through research studies for individuals with DROSHA GPVs.