Review papers with raw data transparency

A. ecDNA landscape in PDAC PDOs

• Using WGS + AmpliconArchitect, the authors classify focal copy-number events into linear, BFB, complex, or circular (ecDNA) amplicons, and find ecDNA present in 12/41 PDOs. AmpliconArchitect ecDNA detection
• MYC is amplified in 11 PDOs, but only 2 PDOs have MYC on ecDNA (ecMYC), while other MYC amplification cases are chromosomal/linear/other categories. MYC ecDNA vs non-ecDNA counts

B. ecDNA drives large cell-to-cell MYC dosage heterogeneity

• Metaphase DNA FISH shows no MYC-positive ecDNAs in icMYC PDOs, while ecMYC PDOs show “ten to hundreds” of MYC-positive ecDNAs per nucleus. FISH metaphase ecMYC vs icMYC
• Interphase FISH signals are confined to spatial nuclear regions with significant short-range clustering, and the two ecDNA-bearing cultures show the greatest variability in MYC copy number and oncogene expression. Interphase clustering and variability
• Importantly, transcriptional output does not scale linearly with ecDNA copy number; ecDNA regulatory landscape (cis elements) modulates expression—for example, the predicted ecDNA in VR06-O lacks PVT1 promoter/first exons, which the authors connect to different MYC expression magnitudes across ecDNA structures. Nonlinear copy-number-to-expression mapping; PVT1 promoter

C. WNT withdrawal selects for MYC dosage strategies and ecDNA dynamics

• In WNT-depleted conditions (-WR), none of the tested PDOs survived continuous passaging, establishing strong environmental hostility. WNT withdrawal extinction (continuous passaging)
• MYC overexpression can eliminate requirement for exogenous WNT agonists (WRi) and also produces insensitivity to C59 (porcupine inhibitor blocking endogenous WNT ligand production), leading the authors to define WR independence. MYC overexpression gives WRi and C59 insensitivity
• When rechallenged with WNT-depleted medium (without passaging), extinction occurred for three cultures, while WRi emerged in the two ecMYC and one icMYC cultures; the kinetics varied but were consistent across replicates for those cultures. Selection outcomes in ecMYC vs icMYC under -WR

D. reversibility + fitness cost of ecDNA burden

• In their culture system, acquisition of WRi is associated with increased per-cell ecDNA content in MYC-bearing ecDNA-driven cells, and mean MYC copy number is tracked by ddPCR and FISH. ecDNA burden rises with WRi; ddPCR support
• Conversely, forced MYC overexpression followed by WR depletion leads to rapid decrease in MYC ecDNA molecules per cell, while MYC expression persists; the authors use this to argue distinct dynamics between MYC levels and ecDNA retention/burden. MYC expression decouples from ecDNA burden
• The paper links high ecDNA burden to increased γH2AX and reduced proliferation; after WNT reintroduction, MYC-high cells reduce fitness (shifting to a non-proliferative fraction), and long-term WNT exposure leads to decreased ecDNA+ fraction and decreased MYC copy number/molecule counts. fitness cost and reversibility after WNT re-addition

E. phenotypic and spatial consequences

• ecMYC accumulation under WRi is associated with morphology shifts (cystic-like → solid/cribriform) that are rapidly reversible when selection pressure is removed; icMYC shows a different behavior under the same logic. Reversible morphology changes tied to ecMYC
• Bulk-like subtype programs: ecDNA accumulation strengthens classical and basal programs in an apparently predictable way across replicates for the ecMYC cultures, with less predictable behavior for icMYC. Replicate predictability differs by ecMYC vs icMYC
• Spatial Xenium + FISH integration indicates that MYC-amplified subdomains show reduced epithelial WNT responsiveness (LGR5 expression), with additional relationships to stromal WNT ligand expression and immune cell territories described by the authors. Spatial mapping of MYC amplification with WNT/immune markers

Strengths (what makes the mechanism credible)

• Multimodal measurement: they combine WGS-based ecDNA structural classification with FISH (metaphase + interphase), ddPCR, and Circle-seq/capillary validation for circularity/breakpoints—reducing the chance that “ecDNA” is a purely computational artifact. Orthogonal validation of ecDNA
• Dynamic causality test: WNT withdrawal followed by re-addition directly probes whether ecDNA burden is maintained by selection and whether state changes are reversible when selective pressures change. Reversibility and selection design
• Regulatory landscape framing: their “nonlinear copy number → expression” argument is consistent with the cis-regulatory differences they discuss (PVT1 promoter presence/absence), offering a mechanistic lever beyond raw copy number. cis regulatory landscape interpretation

Blind spots / uncertainties (what we cannot infer confidently from the excerpt)

• ecDNA ≠ proven necessity in vivo: The excerpt does not provide in vivo animal tumor data demonstrating that eliminating MYC ecDNA (by targeted perturbation) is required for WNT-independent growth, morphology, or spatial WNT/immune territory shifts. The authors provide organoid and spatial FFPE evidence, but causality in living tissue microenvironments remains less directly demonstrated here. Causality gap: in vivo necessity not shown in provided text
• Selection vs adaptation: the paper uses random barcodes and barcode diversity reduction/expansion to help distinguish selection from adaptation, but the excerpt does not allow a complete audit of the quantitative transition probabilities (how many barcodes rose/fell, how strong the effect sizes were) across all cultures. Barcode evidence for selection vs adaptation
• Spatial inference limitations: spatial transcriptomics + FISH integration is persuasive, but transcript-to-genome context is indirect; “MYC amplification” and “WNT responsiveness state” correlation may still reflect microenvironmental co-variation rather than a single mechanistic chain. Correlation vs mechanism in spatial analyses
• Generalizability: ecMYC cases are few (2 MYC ecDNA PDOs), so general mechanistic claims about “MYC ecDNA” may depend heavily on those specific ecDNA structures and their cis-regulatory differences. Small ecMYC count in cohort

What would disprove the main claims? (falsifiability checklist)

• If WNT withdrawal did not select for WRi growth in ecMYC-bearing cultures (or ecDNA burden did not change), then the selection/adaptation model would be weakened. Empirical anchor for falsifiability
• If transcriptional output scaled linearly with ecDNA copy number in these contexts, and structural cis-regulatory differences (e.g., PVT1 promoter presence/absence) did not modulate expression, then the “regulatory landscape” argument would be undermined. Nonlinear scaling and cis modulation claim
• If reducing ecDNA burden without changing MYC protein levels did not change γH2AX/proliferation fractions, then the asserted fitness-cost interpretation could be confounded. Fitness cost tied to ecDNA burden

• Design principle: when assessing oncogene amplification heterogeneity, measure genomic context (ecDNA vs chromosomal) and regulatory landscape, not just copy number or expression. Design principle from results
• Experimental logic: couple environmental niche perturbations to tracked genomic elements (FISH/ddPCR/WGS), and test reversibility to distinguish selection dynamics from one-off adaptations. Selection + reversibility as logic
• Spatial translation: use niche-state markers (here LGR5 for canonical WNT responsiveness) alongside genetic amplification markers to connect genomic mechanisms to tissue ecology. Spatial translation via LGR5