Cancer stem cells have long occupied a special place in tumor biology because of what they can do rather than what they look like. These cells can self-renew indefinitely, regenerate the full diversity of cell types within a tumor, and sustain recurrence after treatment. In EGFR-mutant non-small-cell lung cancer, first-line osimertinib produces deep responses and extends survival, but residual disease frequently persists and can later seed a heterogeneous relapse. The review emphasizes that cancer stem cell identity is best judged by function: durable self-renewal, generation of varied progeny, and long-term tumor propagation are far more informative than any single marker such as ALDH or SOX2. Drug resistance, epithelial-to-mesenchymal transition, and quiescence should all be treated as cancer stem cell-associated features, not as substitutes for proof of stem cell identity.

To bring order to a sprawling literature, the authors introduce a four-level evidence hierarchy. Level A requires only a molecular or phenotypic state linked to stemness, such as a marker signature or a single-cell expression profile. Level B demands direct regenerative evidence, like clonogenic regrowth, sphere or organoid formation, or tumor initiation in a defined population. Level C requires that regenerative function to persist after the drug is withdrawn and the cells recover, ideally tested through repeated challenge. Level D, the most demanding, additionally requires prospective labeling or lineage tracing that connects a defined therapy-persistent population to durable cancer stem cell function. Applying this hierarchy, the authors find that most published studies in EGFR-mutant lung cancer sit at Levels A and B, while direct Level D evidence, showing that a prospectively identified persister lineage later acquires serially propagating stem cell function, remains largely missing.

The drug-tolerant persister cell is central to this story. First described over a decade ago as a reversible, chromatin-mediated state, persister cells capture an early phase of adaptation before stable genetic resistance takes over. In osimertinib-treated models, integrated RNA and ATAC sequencing has revealed dynamic regulatory remodeling with dependencies on the chromatin readers BRD4 and TEAD. Established resistant tumors show a different vulnerability: mSWI/SNF chromatin remodeling complexes are required to maintain resistance-associated programs, and disrupting the subunits SMARCA4 or SMARCA2 can restore drug sensitivity. Other work has shown that the transcription factor ASCL1 promotes osimertinib tolerance only in permissive cellular contexts, and that persister-like states may already be detectable before therapy begins, hinting that some relapse potential is baked in from the start.

DNA and histone modifications add further layers of control. The transcription factor ZNF263 normally recruits the methyltransferase DNMT1 to the EGFR promoter, but it is reduced in persister, resistant, and clinical residual samples; restoring ZNF263 improves responses to EGFR-targeted therapy and limits regrowth. HDAC5, an early osimertinib-responsive gene, is induced shortly after drug exposure, and its depletion increases histone acetylation while reducing the subsequent emergence of resistant cells. The picture around EZH2, the enzyme that deposits the repressive H3K27me3 mark, is more nuanced: its loss lowers H3K27me3 at the MET promoter and facilitates MET-driven resistance, while in other models combined EZH2 and G9a activity promotes tumorsphere formation through SMAD4, ERK, and c-Myc signaling. The relevant question, the authors stress, is whether these enzymes change durable regenerative capacity, not merely whether they change resistance.

RNA epigenetics has emerged as a striking cross-layer mechanism linking chromatin, metabolism, and drug tolerance. In resistant cells, the m6A machinery, composed of the methyltransferase METTL3 and the reader YTHDF2, accelerates degradation of cLMNB1, a transcript whose protein product normally promotes FGFR4 degradation and thereby keeps cells sensitive to osimertinib. An m6A-site-mutated version of cLMNB1 is more stable and more active in resistant models, showing that a single RNA modification can flip a sensitizing circuit into a resistance-enabling one. Separately, METTL3-mediated m6A stabilizes the long noncoding RNA SNHG17, which recruits EZH2 to repress the Hippo pathway gene LATS2 and promotes gefitinib resistance. The enzyme NAT10, whose transcription is driven by p300-mediated H3K27 acetylation, installs ac4C marks that stabilize FATP4 and CPT1A transcripts, supporting fatty acid metabolism and EGFR-TKI resistance. These studies elegantly connect RNA modification with chromatin regulation and metabolism, though they mainly examine established resistant states.

Adaptive signaling pathways overlap heavily with stemness programs and act as effectors of epigenetic competence. YAP and TEAD support apoptosis-resistant persistence, a dependence confirmed by genome-wide CRISPR screens in EGFR-mutant persisters. NOTCH1 and its target genes are elevated in osimertinib persisters, and gamma-secretase inhibition impairs persistence, while Notch3-dependent beta-catenin signaling supports adaptive survival. PAK2-mediated beta-catenin signaling increases cancer stem-like properties in established resistant cells, and protective autophagy maintains SOX2 and ALDH1A1-associated phenotypes during drug stress. The limitation of these studies is temporal: because measurements are made at resistant endpoints, they cannot show when the regenerative phenotype first appeared or whether it arose within a persister lineage. Metabolism and the immune microenvironment extend this framework, with inosine accumulating in resistant cells and residual tumors, where it fuels oxidative phosphorylation and reprograms macrophages toward an immunosuppressive M2-like state.

Perhaps the most clinically provocative idea in the review is that minimal residual disease represents a window in which relapse-competent states are present but not yet irreversibly established. Clinical data support a branching model rather than a single persister-to-stem-cell route: resistance can arise through selection of pre-existing clones or through evolution from drug-tolerant intermediates, and lineage tracing shows that persister lineages carry distinct pretreatment programs and continue to change during therapy. Therapy itself can accelerate genetic evolution, with APOBEC3A-induced mutagenesis and AXL-associated error-prone replication linking cycling persisters to acquired resistance. On the therapeutic side, timing matters: early HDAC5 inhibition may prevent adaptive-state formation, YAP/TEAD targeting suppresses persister survival, and once resistance is established, mSWI/SNF disruption, NAT10-ac4C targeting, or inosine-A2A receptor blockade become relevant. TROP2-directed CAR-T cells can eradicate osimertinib-induced persisters and extend relapse-free survival in vivo, while CD24 blockade enhances macrophage clearance of residual cells, and SOS1 inhibition reduces persister emergence together with tumor-initiating frequency, providing one of the strongest functional bridges toward relapse competence.

The authors close with a concrete experimental roadmap for what would constitute definitive proof of therapy-induced cancer stem cell acquisition. Cells would be barcoded before treatment, persisters identified prospectively during active osimertinib exposure, and matched lineages sampled through acute response, minimal residual disease, drug withdrawal, proliferative recovery, rechallenge, and relapse. Paired single-cell RNA and ATAC sequencing, complemented where warranted by DNA methylation, histone, or RNA-modification profiling, would track regulatory states, while serial organoid regeneration and serial transplantation would test durable function. If perturbing a candidate epigenetic regulator during the persister window allowed cells to recover bulk proliferation but abolished serial tumor-propagating capacity, the experiment would tie a treatment-stage mechanism to later regenerative competence. Until such lineage-resolved studies are done, the review argues, the field should resist equating drug survival with stem cell identity, and should instead focus on identifying which residual states are still reversible, which are being stabilized, and which retain the durable regenerative power to restart the disease.

Subject of Research: Epigenetic regulation of cancer stem cell plasticity and relapse competence in EGFR-mutant lung cancer

Article Title: Epigenetic regulation of cancer stem cell plasticity and relapse competence in EGFR-mutant lung cancer

Article References: Sun, X., Xu, S., & Guo, Q. (2026). Epigenetic regulation of cancer stem cell plasticity and relapse competence in EGFR-mutant lung cancer. Frontiers in Oncology, 16, Article 1981354. https://doi.org/10.3389/fonc.2026.1981354

Image Credits: AI Generated

DOI: 10.3389/fonc.2026.1981354

Keywords: cancer stem cells, drug-tolerant persister cells, EGFR-mutant lung cancer, epigenetic plasticity, osimertinib, minimal residual disease, chromatin remodeling, RNA epigenetics, drug resistance, lineage tracing, tumor relapse, non-small-cell lung cancer