T cell immunotherapies have revolutionized treatment for a subset of cancers. Yet, a major hurdle has been the lack of facile and predicative preclinical animal models that permit dynamic visualization of T cell immune responses at single-cell resolution in vivo. Here, optically clear immunocompromised zebrafish were engrafted with fluorescent-labeled human cancers along with chimeric antigen receptor T (CAR T) cells, bispecific T cell engagers (BiTEs), and antibody peptide epitope conjugates (APECs), allowing real-time single-cell visualization of T cell–based immunotherapies in vivo. This work uncovered important differences in the kinetics of T cell infiltration, tumor cell engagement, and killing between these immunotherapies and established early endpoint analysis to predict therapy responses. We also established EGFR-targeted immunotherapies as a powerful approach to kill rhabdomyosarcoma muscle cancers, providing strong preclinical rationale for assessing a wider array of T cell immunotherapies in this disease.

SUMMARY: Rhabdomyosarcoma (RMS) is a pediatric tumor that resembles undifferentiated muscle cells; yet the extent to which cell state heterogeneity and molecular features are shared with human development have not been fully ascribed. Here, we report a single-cell/nucleus RNA sequencing atlas derived from 72 datasets that includes patient tumors, patient-derived xenografts, primary in vitro cultures, and established cell lines. We report four dominant muscle-lineage cell states in RMS: progenitors, proliferative, differentiated, and ground cells. We stratify these RMS cells along the continuum of human muscle development and show that RMS cells share expression patterns with fetal/embryonal myogenic precursors rather than postnatal satellite cells. Indeed, fusion-negative RMS (FN-RMS) have a discrete stem cell hierarchy that faithfully recapitulates fetal muscle development. We also identify therapy-resistant FN-RMS progenitor cells that share transcriptomic similarity with bipotent skeletal mesenchymal cells, while a subset of fusion-positive (FP) RMS have tumor-acquired cells states, including a neuronal cell state, that are not found in development. Chemotherapy induced upregulation of progenitor signatures in FN-RMS while the neuronal gene programs were retained after therapy in FP-RMS. Taken together, this work identifies new cell state heterogeneity including unique treatment-resistant and tumor-acquired cell states that differ across RMS subtypes.

Despite comprehensive and multi-modal therapy, outcomes for children and adolescents with rhabdomyosarcoma (RMS) have plateaued over the past four decades. This is not for a lack of progress in the basic and translational studies of RMS. Indeed, advances in animal models and/or patient tissue sample acquisition and analysis have improved our understanding of RMS biology. Large-scale sequencing efforts have generated transcriptomic, genomic, and epigenomic datasets that highlight the heterogeneity of RMS and have the potential to improve prognostication and the application of precision medicine in patients with RMS. However, few of these discoveries have been clinically translated, and limitations to the accessibility, uniformity, and application of these new models and datasets hinder their utility. Here, we discuss how advances in understanding RMS biology, optimization of preclinical models, and strategies for translating basic science discoveries to the clinic can potentially improve outcomes for patients with RMS.