When a cell doubles its entire genome — whether a plant breeder's tetraploid rice, a megakaryocyte replicating DNA without dividing, or a tumor that has undergone whole-genome doubling (WGD) — it becomes giant, and its transcriptome shifts. At the same time, animal cells possess a built-in self-destruct program called pyroptosis: sensor proteins detect danger, activate caspase enzymes that act as molecular scissors, and these scissors liberate the N-terminal fragment of a gasdermin protein, which punches symmetric pores of roughly 15–32 nanometers across the plasma membrane Microglia Pyroptosis: A Candidate Target for Neurological Diseases Treatment. Water rushes in through the pores, the cell swells and bursts, dumping inflammatory signals and DNA into its surroundings. These two phenomena are almost never studied together, and their intersection hides a measurement problem that matters clinically: colorectal tumors with WGD shed 53% more detectable circulating tumor DNA (ctDNA) into blood Impact of Whole Genome Doubling on Detection of Circulating Tumor DNA in Colorectal Cancer, and nobody knows whether that is because each dying cell simply carries twice the DNA, or because genome doubling actively rewires how cells die.
The core problem is that RNA sequencing — our only genome-wide readout of these programs — cannot, as usually analyzed, distinguish three things in a polyploid cell: passive gene-dosage scaling (more copies, more transcripts), active regulation of the death machinery, and changes in execution, which is post-translational (a gasdermin transcript is useless until a caspase cleaves the inhibitory domain off the protein). This proposal builds a dosage-calibrated, geometry-aware way to read the inflammatory death program in polyploid cells from RNA-seq, and then tests its predictions with cleavage and pore-formation assays in polyploidized human cell lines.
Core Claim
The claim has two falsifiable parts. First, in genome-doubled cells the apparent expression of pyroptosis genes (NLRP3, CASP1, GSDMD, GSDME, CASP3) is inflated by dosage, but once each gene's fold-change is calibrated against the genome-wide ploidy-driven shift, a residual, actively regulated signature remains — and its direction (suppression or amplification of lytic death) is the biologically meaningful signal that current models throw away. Second, that residual signature, not passive DNA mass, explains the stage-dependent ctDNA shedding advantage of WGD+ colorectal tumors: passive stoichiometry cannot explain why the advantage collapses in Stage III disease (OR = 0.83, 95% CI: 0.44–1.53) Impact of Whole Genome Doubling on Detection of Circulating Tumor DNA in Colorectal Cancer, while an active death-mode shift — for example, stage-dependent epigenetic silencing of gasdermin genes, documented for GSDME in triple-negative breast cancer Pyroptosis‐Inducing Platinum(IV) Prodrugs via GSDME Pathway for Chemoimmunotherapy and Metastasis Inhibition in Triple‐Negative Breast Cancer — can.
Research Program
The program has three connected components. The first is a re-analysis of existing public data that has never been queried for this question. The Danish colorectal cohort (833 patients, 701 after quality control) paired WES-based WGD calls with ctDNA detection but performed no RNA sequencing Impact of Whole Genome Doubling on Detection of Circulating Tumor DNA in Colorectal Cancer; we will instead call WGD in TCGA colorectal and pan-cancer samples with the same ASCAT definition (>50% of the genome with major allele copy number ≥ 2) and pair those calls with TCGA RNA-seq. Separately, the wedelolactone study deposited bulk RNA-seq of Meg-01 and K562 cells driven into endomitotic polyploidization (≥ 8N) under accession PRJNA1162465 Regulating mitochondrial oxidative phosphorylation and MAPK signaling: wedelolactone as a novel therapeutic for radiation-induced thrombocytopenia — a homogeneous, isogenic polyploidization time course that is ideal for isolating dosage effects, yet it was analyzed only for metabolic pathways and never for the inflammasome–gasdermin axis, despite the fact that the same drug treatment elevates reactive oxygen species, the known trigger of NLRP3 assembly Rab32 facilitates Schwann cell pyroptosis in rats following peripheral nerve injury by elevating ROS levels. As a cross-kingdom check that polyploidy genuinely perturbs programmed cell death timing, we will analyze the neo-tetraploid rice anther dataset PRJCA058658, where delayed tapetal cell death cut seed-setting from 78.10% to roughly 37–39% and abscisic acid rescue restored it to about 60% <i>HSP101</i> ‐encoding NEO‐TETRAPLOID RICE FERTILITY GENE 1 regulates tapetum development through interaction with SAPK2 in polyploid rice (HSP101 ‐encoding NEO‐TETRAPLOID RICE FERTILITY GENE 1 regulates tapetum development through interaction with SAPK2 in polyploid rice); this motivates, and does not overlap with, our mammalian inflammatory-death question, since tapetal death is plant autolytic PCD with no gasdermin execution.
The second component is the calibration itself. For each death gene $g$ we define a dosage deviation score
$$d_g \;=\; \log_2\!\frac{\mu_g^{+}}{\mu_g^{-}} \;-\; \operatorname{median}_{j}\,\log_2\!\frac{\mu_j^{+}}{\mu_j^{-}},$$
where $\mu_g^{+}$ and $\mu_g^{-}$ are mean expression of gene $g$ in WGD+ and WGD− tumors and the median runs over all expressed genes $j$, giving the genome-wide ploidy-driven shift. Under pure dosage, $d_g \approx 0$ for every gene; a significantly negative $d_g$ for gasdermin genes would mean active suppression of lytic death despite doubled copies. We then fold this into a predictive signature — a ploidy-aware lytic-death score — that can be applied to any bulk RNA-seq profile, and we will test whether adding an explicit ploidy-conditioning input improves over strong existing transcriptomic models (see Baselines).
The third component is functional validation, because RNA-seq alone cannot see cleavage. We will reproduce wedelolactone-induced polyploidization in Meg-01 and K562 cells (flow-cytometric ploidy bins of 2N, 4N, ≥ 8N, plus CD41/CD61 markers as in the original study Regulating mitochondrial oxidative phosphorylation and MAPK signaling: wedelolactone as a novel therapeutic for radiation-induced thrombocytopenia), challenge the sorted fractions with sterile inflammasome triggers such as monosodium urate crystals, and measure full-length versus N-terminal gasdermin fragments by western blot, pore permeability by calcein-AM/propidium iodide, cytokine release by IL-1β and IL-18 ELISA, and time-to-rupture by live-cell imaging of ballooning — the standard toolkit of the field Microglia Pyroptosis: A Candidate Target for Neurological Diseases Treatment, NLRP3 inflammasome in endothelial dysfunction. This directly tests an unstudied biophysical question: giant polyploid cells have a different surface-area-to-volume ratio, so the number of gasdermin pores needed to reach lethal osmotic tension should change with ploidy — possibly explaining how megakaryocytes survive massive ROS bursts during endomitosis without lysing, a question the megakaryocyte literature leaves open Insights into Regulatory Factors in Megakaryocyte Development and Function: Basic Mechanisms and Potential Targets.
Baselines
The naive reference point is the implicit model behind the Danish ctDNA study: shedding scales passively with DNA mass per cell, and any RNA-seq difference between WGD+ and WGD− tumors is uniform dosage Impact of Whole Genome Doubling on Detection of Circulating Tumor DNA in Colorectal Cancer. Concretely this is the prediction $d_g = 0$ for all death genes and linearly dosage-scaled cleavage in the cell lines; we beat it by showing statistically significant, reproducible deviation that tracks functional cleavage differences.
The strongest existing competitor on the pyroptosis-from-RNA-seq question is the nine-gene pyroptosis target signature and its BFReg-NN model, which achieved a concordance index of 0.90 for predicting drug synergy in the 360-patient FUSCC triple-negative breast cancer cohort AI-powered omics-based drug pair discovery for pyroptosis therapy targeting triple-negative breast cancer. That signature uses raw expression of pyroptosis genes with no ploidy awareness, and its authors explicitly did not examine genome-doubled subpopulations. We will run their nine-gene signature on our WGD-stratified TCGA splits with the same cross-validation folds, sample sizes, and tuning budget as our calibrated score, using cleavage-validated cell-line labels projected into tumor-deconvolved space as the evaluation target. Beating it means higher discrimination on lytic-death labels and, more importantly, correct attribution: where the raw signature calls a WGD+ tumor "pyroptosis-high," our score must decide whether that is dosage or regulation, and the cell-line cleavage assays must confirm the call.
A second baseline is BulkRNABert, a Transformer language model over binned TPM values, which reports a weighted F1-score of 0.942 across 33 TCGA cancer types and a pan-cancer survival C-index of 0.765 BulkRNABert: Cancer prognosis from bulk RNA-seq based language models, with public code and weights multiomics-open-research. We will fine-tune it with the same parameter-efficient $IA^3$ scheme on our labels under an identical compute budget; this baseline tests whether representation power alone absorbs what our calibration ingredient provides. Finally, the key ablation is our own score computed without the dosage-normalization step, on raw TPM — this isolates ploidy calibration as the sole added ingredient, since the gene panel, data, and training procedure are otherwise identical.
Why not just…?
Why not just assume passive DNA dosage explains the ctDNA result? This is the explanation the Danish authors themselves favor, and it is genuinely parsimonious. But it makes a prediction their own data refute: if shedding were stoichiometric, the WGD advantage should grow or persist in Stage III tumors, where tumor mass and vascular access are greatest; instead the association vanishes (OR = 0.83, 95% CI: 0.44–1.53) Impact of Whole Genome Doubling on Detection of Circulating Tumor DNA in Colorectal Cancer. Passive dosage also cannot explain how a giant megakaryocyte survives ROS levels that trigger pyroptosis in diploid cells Rab32 facilitates Schwann cell pyroptosis in rats following peripheral nerve injury by elevating ROS levels. Still, we take the skeptic seriously: if our cell-line experiments show gasdermin cleavage and lysis scaling purely with ploidy at matched stimulus, and $d_g \approx 0$ across the panel, we will report that the passive model holds and that the Stage III collapse is due to something else (immune clearance), which our TCGA stage stratification can also interrogate.
Why not just use single-cell RNA-seq instead of bulk with calibration? For heterogeneous tumor tissue this is a real threat, since bulk averages over polyploid tumor cells, stroma, and infiltrating immune cells, and single-cell methods demonstrably resolve cell diversity that bulk obscures Single-Cell RNA Sequencing Outperforms Single-Nucleus RNA Sequencing in Analyzing Pancreatic Cell Diversity and Gene Expression in Goats. Our answer is practical: the dosage question is cleanly answerable in isogenic homogeneous cultures, where bulk RNA-seq already captured endopolyploid shifts in the wedelolactone dataset Regulating mitochondrial oxidative phosphorylation and MAPK signaling: wedelolactone as a novel therapeutic for radiation-induced thrombocytopenia, and in tumors we combine deconvolution with the fact that dosage calibration needs only the WGD− matched comparison, not per-cell resolution. Where ambiguity remains, a small single-nucleus pilot on high-ambiguity TCGA-like samples will settle it, and we budget for that rather than argue around it.
Why not just measure gasdermin cleavage directly and skip the transcriptomics? Cleavage westerns and pore-permeability assays are the ground truth, and we use them as such — but they are low-throughput end-point readouts on a handful of proteins. They cannot define the genome-wide dosage baseline against which "deviation" is even meaningful, they cannot be run on the thousands of archival TCGA and FUSCC profiles that only have RNA, and they yield no transferable predictive signature for clinical cohorts. Conversely, RNA-only inference of pyroptosis is exactly the trap Gap-level reasoning warns about: dormant full-length gasdermin protein means high mRNA does not equal death Microglia Pyroptosis: A Candidate Target for Neurological Diseases Treatment, Pyroptosis‐Inducing Platinum(IV) Prodrugs via GSDME Pathway for Chemoimmunotherapy and Metastasis Inhibition in Triple‐Negative Breast Cancer. The design deliberately couples the two: RNA defines the candidate space, cleavage assays provide the labels and the arbitration.
Proposed Experiments and Decisive Checks
Experiment one computes $d_g$ for the full inflammatory death panel in TCGA colorectal cancer, stratified by stage, with WGD called by ASCAT under the published definition. The decisive readout is whether gasdermin and inflammasome genes deviate from the dosage median in early-stage WGD+ tumors and whether that deviation reverses or the genes are silenced in late-stage tumors — the pattern that would explain the Stage I (OR = 2.44) to Stage III (OR = 0.83) collapse reported in the Danish cohort Impact of Whole Genome Doubling on Detection of Circulating Tumor DNA in Colorectal Cancer. We must be honest about a limit: no cohort on file pairs RNA-seq with ctDNA detection — no reported result found on Lacuna for RNA-seq in the Danish cohort — so the tumor-leg test is mechanistic consistency, not direct ctDNA prediction; a prospective matched collection is the follow-up this work would justify.
Experiment two re-analyzes PRJNA1162465 for the inflammasome–gasdermin axis across wedelolactone-induced polyploidization of Meg-01 and K562 cells, testing whether death-machinery transcripts stay flat, scale with dosage, or are actively suppressed as cells climb from 2N to ≥ 8N Regulating mitochondrial oxidative phosphorylation and MAPK signaling: wedelolactone as a novel therapeutic for radiation-induced thrombocytopenia.
Experiment three is the functional core: sorted 2N, 4N, and ≥ 8N fractions challenged with sterile inflammasome activators, read out by gasdermin cleavage, calcein-AM/propidium iodide permeability, IL-1β/IL-18 ELISA, and live-cell time-to-rupture, with modality attributed pharmacologically — disulfiram to block GSDMD-N oligomerization Disulfiram alleviates pristane-induced lupus via inhibiting GSDMD-mediated pyroptosis, VX-765 for caspase-1, and Z-DEVD-FMK for caspase-3, whose blockade was shown to force cells back into silent apoptosis Pyroptosis‐Inducing Platinum(IV) Prodrugs via GSDME Pathway for Chemoimmunotherapy and Metastasis Inhibition in Triple‐Negative Breast Cancer. This inhibitor logic also rules out the secondary-necrosis alternative, since gasdermin cleavage precedes and is required for the lytic, cytokine-releasing phenotype Microglia Pyroptosis: A Candidate Target for Neurological Diseases Treatment. The decisive quantities are pore-count-normalized rupture time and cytokine release per cell as functions of ploidy.
Experiment four builds the ploidy-aware lytic-death score, trains it on the cell-line labels plus TCGA, and benchmarks it on the same splits against the BFReg-NN nine-gene signature (whose reported drug-synergy C-index of 0.90 AI-powered omics-based drug pair discovery for pyroptosis therapy targeting triple-negative breast cancer sets the bar for pyroptosis-signature models) and a BulkRNABert fine-tune (reported pan-cancer survival C-index of 0.765 BulkRNABert: Cancer prognosis from bulk RNA-seq based language models), with survival stratification tested on the 360-patient FUSCC cohort as an external check.
The program is falsifiable at defined points: uniform $d_g \approx 0$ plus dosage-linear cleavage would vindicate the passive model; stage-independent gasdermin suppression would redirect the ctDNA explanation toward immune clearance; and confirmation would deliver both a corrected way to read inflammatory death in the substantial fraction of human tumors that are genome-doubled and a mechanistic account of how nature's giant polyploid cells avoid blowing themselves up.
References
- Microglia Pyroptosis: A Candidate Target for Neurological Diseases Treatment
- Impact of Whole Genome Doubling on Detection of Circulating Tumor DNA in Colorectal Cancer
- Pyroptosis‐Inducing Platinum(IV) Prodrugs via GSDME Pathway for Chemoimmunotherapy and Metastasis Inhibition in Triple‐Negative Breast Cancer
- Regulating mitochondrial oxidative phosphorylation and MAPK signaling: wedelolactone as a novel therapeutic for radiation-induced thrombocytopenia
- Rab32 facilitates Schwann cell pyroptosis in rats following peripheral nerve injury by elevating ROS levels
- HSP101 ‐encoding NEO‐TETRAPLOID RICE FERTILITY GENE 1 regulates tapetum development through interaction with SAPK2 in polyploid rice
- NLRP3 inflammasome in endothelial dysfunction
- Insights into Regulatory Factors in Megakaryocyte Development and Function: Basic Mechanisms and Potential Targets
- AI-powered omics-based drug pair discovery for pyroptosis therapy targeting triple-negative breast cancer
- BulkRNABert: Cancer prognosis from bulk RNA-seq based language models
- Single-Cell RNA Sequencing Outperforms Single-Nucleus RNA Sequencing in Analyzing Pancreatic Cell Diversity and Gene Expression in Goats
- Disulfiram alleviates pristane-induced lupus via inhibiting GSDMD-mediated pyroptosis