ISWI double-knockout signatures stratify PAX3/7-FOXO1 fusion-positive vs fusion-negative rhabdomyosarcoma

Background: Sarcoma fusion oncogenes drive distinct transcriptional programs that are aligned with chromatin landscapes shaped by ATP-dependent remodeling complexes. In rhabdomyosarcoma (RMS), fusion-positive (FP) tumors harbor PAX3/7-FOXO1 fusions, whereas fusion-negative (FN) tumors lack FOXO1 fusions. We aimed to integrate fusion transcript detection with gene program scoring to quantify how fusion status relates to chromatin remodeler module activity, and to test whether an experimentally defined ISWI perturbation signature is enriched in fusion-positive disease. Methods: We analyzed bulk transcriptomes from the public RMS cohort GSE108022 (Tumor n=101; Normal skeletal muscle n=5; FN n=66; FP n=35). Fusion status and fusion class (FP vs FN) were determined from RNA-seq using fusion transcript detection and dataset annotations. We computed ssGSEA/GSVA enrichment scores for a compact chromatin remodeling panel capturing core complex programs (e.g., BAF, ncBAF, NuRD, ISWI) and ISWI/SMARCA-linked target signatures, including two ISWI double-knockout (DKO) response programs (Smarca1/Smarca5; ISWI ATPase subunits SMARCA1/SNF2L and SMARCA5/SNF2H) derived from mouse DKO vs WT differential expression and mapped to human orthologs (ISWI_DKO_targets_UP/DOWN). Group differences were assessed using two-sided Wilcoxon rank-sum tests with Benjamini-Hochberg correction (Tumor vs Normal; FP vs FN within tumors). We additionally evaluated fusion-class discrimination using a module-only logistic regression classifier with repeated 5-fold cross-validation and AUROC. Results: Compared with normal muscle, RMS tumors showed significantly higher enrichment of multiple chromatin remodeling programs, including BAF_core, NuRD_core, ncBAF, and ISWI_core (all FDR 6.500357e-04). ISWI_DKO_targets_UP was lower in tumors than normal muscle (tumor median z=-0.008 vs normal median z=1.569; FDR 1.936228e-03). ISWI_DKO_targets_DOWN was higher in tumors than normal muscle but showed only a borderline trend after correction (tumor median z=0.108 vs normal median z=-0.592; FDR 5.148731e-02). Within tumors, fusion class strongly stratified the mechanistic ISWI dependence program: ISWI_DKO_targets_DOWN was markedly higher in FP than FN (p=2.521292e-07; FDR=2.521292e-06; FP median z=0.775; FN median z=-0.422; delta=1.197). Additional FP-FN differences included SMARCA1_targets_UP (FDR 2.206301e-02). An ISWI+SMARCA1/5 module-only logistic regression model discriminated FP from FN tumors with AUROC=0.876±0.012. Conclusions: A minimal ssGSEA panel integrating chromatin remodeling complex cores with an ortholog-mapped ISWI DKO program identifies a fusion-linked chromatin remodeling state in RMS. ISWI_DKO_targets_DOWN shows the strongest FP-FN separation at the program level and supports compact, reproducible fusion-class stratification, motivating broader validation and functional follow-up of ISWI-associated dependencies in fusion-driven RMS. This chromatin-centric mapping strategy provides a practical route to develop candidate program-level biomarkers and to prioritize testable combination hypotheses linking fusion-driven transcriptional output with chromatin remodeling programs implicated in cell-state plasticity.