The resistance mechanism your organoid assay cannot see

What the work claims

Huang and colleagues claim that sulfide:quinone oxidoreductase (SQOR), an inner-mitochondrial-membrane flavoprotein that oxidizes hydrogen sulfide and passes the electrons to coenzyme Q (CoQ), is a hypoxia-inducible defense against ferroptosis in PDAC, and that this defense contributes to resistance against gemcitabine, the chemotherapy backbone for this disease.¹ The claim is assembled from a 102-patient tissue microarray, two KRAS-mutant cell lines manipulated with knockdown, overexpression, and a catalytically dead mutant, pathway-epistasis experiments against the three canonical ferroptosis-defense branches, death-mode rescue assays, eight patient-derived organoid (PDO) lines, and two xenograft cohorts.

This is a primary experimental study with an exploratory clinical layer. Its boldest element is not the association but the mechanism: that sulfide oxidation feeds electrons into the mitochondrial CoQ pool to keep it reduced (as CoQH2), and that this is partially nonredundant with the GPX4, FSP1, and DHODH systems that the field already treats as the ferroptosis defense network.

How it works

Ferroptosis is cell death driven by runaway oxidation of polyunsaturated phospholipids in membranes, executed when iron-dependent peroxidation outruns repair. Tumors defend against it through several branches: GPX4, an enzyme that directly reduces phospholipid hydroperoxides using glutathione; FSP1, which regenerates CoQH2 at non-mitochondrial membranes; and DHODH, a mitochondrial enzyme that supplies a CoQ-linked defense. SQOR adds a candidate fourth input, donating electrons from hydrogen sulfide (H2S) into the same CoQ pool.

The evidence chain runs as follows. In the patient cohort, SQOR protein staining correlated moderately with the hypoxia markers CAIX (Spearman rho 0.473) and HIF-1alpha (rho 0.438) and inversely with 4-HNE, a lipid-peroxidation marker (rho -0.465), across 102 patients analyzed one-per-patient rather than one-per-core.¹ In cells at 1 percent oxygen, SQOR rose within 24 hours and the reduced CoQ fraction climbed about 1.6-fold. Knockdown or a catalytically dead point mutant reversed the shift and increased both whole-cell and mitochondrial lipid reactive oxygen species; the dead mutant phenocopied knockdown, so catalysis rather than protein mass does the work. H2S donors (GYY4137, NaHS) raised the reduced CoQ fraction and lowered lipid ROS, while inhibiting or silencing the H2S-producing enzymes CBS and CSE blunted the hypoxia response; selenium-route perturbations did not reach significance. Epistasis placed SQOR as partially independent of the known branches: knockdown synergized with GPX4 inhibition (combination index 0.68 at ED50), FSP1 inhibition (0.74), and DHODH inhibition (0.52 with brequinar, 0.57 with BAY 2402234), and SQOR overexpression rescued only about 18 percent of the lipid-ROS elevation in a DHODH-deficient background versus about 40 percent in GPX4- or FSP1-deficient ones, consistent with convergence on mitochondrial CoQ.

On the therapeutic side, gemcitabine-resistant cells carried about 2.8-fold more SQOR protein, a reduced CoQ fraction of about 49 percent versus 27 percent in parental cells, and half the baseline lipid ROS. A putative SQOR-directed compound partially resensitized both resistant lines (gemcitabine IC50 from 2.437 to 0.812 micromolar in PANC-1-GEM-R, a 13.2-fold parental gap narrowed to 4.4-fold), and adding a ferroptosis inducer narrowed it further to 2.0-fold. In eight PDO lines stratified by parent-tumor SQOR staining, the compound plus gemcitabine cut viability by 46 to 52 percentage points in the four SQOR-high lines (mean -49.0 points) but only by -11 to +3 points in the four SQOR-low lines (exploratory treatment-by-subgroup interaction, P = 0.0036). In xenografts, adding the compound to gemcitabine reduced day-28 tumor volume from 892 to 472 cubic millimeters and extended median survival from 35 to 49 days.

Where a skeptic should push

The single most load-bearing assumption is that the compound used in every therapeutic experiment actually works through SQOR. It is described only as a putative SQOR-directed agent, and the paper itself supplies the reason for doubt: the best-characterized SQOR inhibitor in the literature, STI1, shows off-target mitochondrial effects at concentrations only 350-fold above its enzymatic IC50, and the present study measured no direct SQOR engagement in cells, organoids, or tumors. The xenograft efficacy endpoints, by the authors' own statement, do not establish target engagement or ferroptotic death in vivo. Until the compound's selectivity and target engagement are nailed down, the entire therapeutic half of the paper is mechanism-consistent but not mechanism-proven.

Second, the clinical layer is association only, and the paper is unusually honest about it. The correlations are moderate, the response endpoint mixes RECIST imaging with a study-specific CA19-9 threshold, the SQOR cutoff of 156 was selected and evaluated in the same cohort (SQOR-high patients had median overall survival of 14.2 months versus 23.8 months, but the cutoff carries selection optimism), and the tissue data do not establish that hypoxia lowers lipid peroxidation through SQOR in patients.

Third, generalization. Mechanism rests on two KRAS-mutant cell lines; the PDO experiment has four lines per arm with a descriptive median split; and no independent cohort validates anything. The authors also disclose that generative AI was used for language editing, which bears on prose, not data, but the absence of declared funding and the modest scale of the translational stages both argue for reading this as hypothesis-generating preclinical work.

Hypoxia, the hidden organoid assay variable

The non-obvious implication cuts at a default of the field: nearly every PDO drug-sensitivity assay is run at atmospheric oxygen, around 21 percent, while the tumor regions this paper implicates sit near 1 percent. SQOR induction, the CoQH2 shift, and the resulting ferroptosis threshold are all oxygen-gated. A PDO grown in air therefore under-expresses the very defense the study says drives resistance, and it will mis-rank ferroptosis-directed combinations, reading a SQOR-high tumor as more sensitive than its in-vivo behavior warrants. This is the same class of problem as medium formulation quietly rewiring drug response, but it is worse, because oxygen is rarely even recorded as assay metadata.

The opportunity is concrete. The study's own eight-line experiment is a template for biomarker-stratified PDO pharmacotyping: stain the parent tumor for SQOR (and, sensibly, CAIX), split lines by the marker, and read combination benefit as an interaction rather than a population average. For ferroptosis-sensitizer programs, which have repeatedly failed to convert single-agent cell killing into patient benefit, a hypoxia-matched or at minimum hypoxia-documented PDO layer could filter out compounds whose window exists only in air. The threat is symmetrical and commercial: if vendors and core facilities standardize on normoxic PDO response as the predictive endpoint without an oxygen annotation, biomarker-guided combinations can be discarded for looking ineffective in exactly the conditions where the biology is switched off. A credible organoid drug-response package should now carry oxygen tension the way it carries passage number and mutation status; a result without it is, for this pathway class, close to uninterpretable.

The bottom line

Established, within the tested models: hypoxia induces SQOR catalytic activity, SQOR keeps mitochondrial CoQ reduced, and that protection raises the ferroptosis threshold in a partially nonredundant way alongside GPX4, FSP1, and DHODH. Suggestive, not proven: that this axis drives gemcitabine resistance in patients or that the unnamed putative inhibitor acts through SQOR at all. The confirmatory experiments are a characterized, selective SQOR inhibitor with demonstrated target engagement in organoids, and a prospective PDO series cultured under documented oxygen with a prespecified SQOR cutoff. For anyone running pancreatic tumor organoids in drug screens, the immediate action costs nothing: record the oxygen, and treat a ferroptosis readout taken in air as a lower bound, not a measurement.