Acquired resistance of cancer arises from continual evolution induced by therapeutic pressure. Breaking conventional evolutionary barriers, extrachromosomal DNA (ecDNA) recently emerged as a powerful driver of tumor adaptability and therapeutic failure. Studying ecDNA-driven evolution has largely focused on ecDNA copy number dynamics. However, this perspective overlooks the adaptive potential of ecDNA structural changes. A key mechanism of structural adaptation is ecDNA reintegration into chromosomes to form homogeneously staining regions (HSRs), elongated stretches of chromosomes that store oncogene copies. Prevalent in resistant malignancies, ecDNA-HSR transition is increasingly suspected to underlie therapeutic evasion. Yet how ecDNA reintegration occurs, how it varies across treatment conditions or cell types, and how it enables adaptation and resistance remain unclear. Addressing these gaps will greatly improve our ability to predict resistance progression and counteract ecDNA-driven malignancies, and this requires systematic characterization of ecDNA and HSRs.To fill these gaps, this work provides a systematic, high-resolution characterization of ecDNA drug responses using state-of-the-art single-cell multi-omics sequencing, imaging, and machine learning. We demonstrate that during drug exposure, ecDNA produces distinct adaptive responses in a cell population through complex genomic alterations. First, we consistently see significant ecDNA reduction in the cells that survived treatments. This is tightly correlated with gene expression changes in the same single cells, evidence that ecDNA is subject to selection. Intriguingly, imaging results have uncovered a previously unrecognized type of ecDNA that may represent an intermediate state bridging ecDNA-HSR transition. This observation challenges the traditional two-state view of ecDNA reintegration, suggesting instead a dynamic continuum that may underlie graded adaptation. To dissect this continuum at the functional levels, we are measuring ecDNA, intermediate, HSRs, RNA, and protein configurations in single cells across diverse adaptation models.Integrating innovative single-cell technology development with mechanistic discovery, our findings underscore the remarkable plasticity of ecDNA implicated in resistance acquisition. We show that rather than solely relying on pre-existing pro-survival traits, cancer cells may additionally exploit ecDNA to fast-track genomic changes that enable a wider range of adaptation. Such genomic plasticity may explain the quick relapse after selective therapies, offering new insights into actionable biomarkers that can reverse ecDNA-driven resistance.