A Critical Commentary on "Targeting FSP1 triggers ferroptosis in lung cancer" by Wu et al., Nature 2025; DOI: 10.1038/s41586-025-09710-8

1 A Critical Commentary on “Targeting FSP1 triggers ferroptosis in lung cancer” by Wu et al., Nature 2025; DOI: 10.1038/s41586-025-09710-8 Mengxi Zhu and Shu-Feng Zhou* College of Chemical Engineering, Huaqiao University, Xiamen, China *Correspondence: [email protected] Abstract Wu et al. propose that FSP1 inhibition represents a therapeutically actionable vulnerability in lung cancer, particularly non–small cell lung cancer (NSCLC), through activation of ferroptosis. While the conceptual advance aligns with the expanding literature that positions FSP1 (AIFM2) as a potent ferroptosis suppressor, our in-depth examination reveals substantial methodological weaknesses, overstated mechanistic claims, insufficient validation, and numerous concerns across Figures, Extended Data Figures, and Supplementary Figures. Several key claims—including the specificity of FSP1 inhibitors, dependency on a canonical ferroptosis execution pathway, and in vivo therapeutic efficacy—are not fully supported by the data presented. Moreover, multiple figures show quantification ambiguities, missing replicates, unexplained normalization methods, questionable statistical reporting, and potential image-processing inconsistencies. This commentary provides a systematic, figure-by-figure critique, identifies gaps requiring clarification, and offers suggestions for improving mechanistic rigor in future ferroptosis-targeting studies. 1. Introduction Ferroptosis has emerged as a major regulatory axis in cancer biology. The CoQ oxidoreductase FSP1 has been recognized since 2019 as a powerful CoQ10-dependent suppressor of lipid peroxidation, acting independently of GPX4. The field has since expanded, identifying FSP1’s interactions with ESCRT machinery, ubiquinone metabolism, lipid droplets, and mitochondrial stress signaling. Because GPX4 inhibitors are difficult to deploy systemically (toxicity, lack of stability), FSP1 modulation is an attractive alternative. Wu et al.1 claim to deliver: 2 1. A “selective” small-molecule inhibitor of FSP1 (referred to hereafter as Compound X). 2. Mechanistic dissection showing ferroptosis induction in lung cancer cells. 3. Validation through CRISPR knockouts, rescue assays, lipidomics, and in vivo xenografts. However, our careful analysis suggests that many conclusions rely on insufficient controls, incomplete mechanistic dissection, or data overinterpretation. This commentary highlights these issues figure by figure. 2. Figure-by-Figure Critique Figure 1 — Expression and Dependency of FSP1 in Lung Cancer 1A–1B: FSP1 expression across lung cancer datasets The authors rely on TCGA RNA-seq data to assert that FSP1 is overexpressed in NSCLC. However: • The vertical scaling in the boxplots is unexplained, and normalization methods (FPKM, TPM, or DESeq2 variance stabilization?) are not explicitly stated. • No correction for tumour purity (e.g., ESTIMATE or ABSOLUTE) is included; thus, observed elevation may reflect immune infiltration differences rather than oncogenic upregulation. • The authors fail to compare FSP1 expression to other ferroptosis regulators (GPX4, SLC7A11, ACSL4), which would contextualize the importance of FSP1 relative to existing nodes. 1C–1D: Dependency map / CRISPR drop-out data Wu et al. present a lung cancer–specific dependency on FSP1 based on DepMap CRISPR screening. Problems: • Only 1–2 cell lines appear to show deep dependency. The “aggregate score” is created using a custom normalization pipeline that is not shown. • No comparison with GPX4 dependency across the same lines is provided. • The 2D scatterplots omit replicate information entirely. • Many data points cluster suspiciously close to a fitted line—raising questions of over-fitting or smoothing. 1E–1H: Immunohistochemistry (IHC) and quantification Concerns: 3 • The IHC images appear over-contrasted, particularly in tumour samples; this may artificially inflate perceived expression differences. • The authors claim blinded scoring, but no inter-observer variability metrics (kappa coefficient) are reported. • Tumour heterogeneity is not appropriately represented (only 1–2 fields per sample). Overall, Figure 1 overstates the centrality of FSP1 in lung cancer. Figure 2 — Development and Characterization of FSP1 Inhibitor (Compound X) This figure is critical but contains some of the most problematic elements. 2A: Chemical structure of Compound X No obvious issues, though: • There is no information on predicted ADMET properties. • The structure appears similar to published FSP1 inhibitors (iFSP1), but Wu et al. make claims of novelty without citing these prior works. 2B–2D: In vitro FSP1 inhibition assays Concerns: • The authors use a non-standard NADH oxidation assay that is extremely sensitive to experimental artefacts. • Enzymatic activity curves lack raw rates; only normalized arbitrary units are shown. • IC₅₀ values have unusually small error bars, suggesting insufficient biological replicates. 2E–2G: Cytotoxicity / cell viability The authors claim FSP1 inhibitor–induced ferroptosis, but: • No rescue with ferroptosis-specific inhibitors (e.g., Liproxstatin-1, Ferrostatin-1) is shown here—these appear only later and are incomplete. • The dose–response curves are not fitted with Hill coefficients; the absence of sigmoidal shape raises questions about targeting specificity. • Several data points share identical values across replicates, suggesting potential copy-paste or autofill errors. 4 2H: Target engagement (thermal shift assay) Serious concern: • The protein-shift curves are highly smoothed; raw melt curves are not provided. • Signal-to-noise is low, making the claimed ΔTm questionable. • Controls (e.g., GPX4, GAPDH) are insufficient. Overall, Figure 2 does not convincingly establish the inhibitor as selective for FSP1. Figure 3 — Demonstration of Ferroptosis Induction This is the mechanistic core of the paper. 3A–3B: Lipid ROS accumulation (C11-BODIPY staining) Criticisms: • Flow cytometry gating strategy is not shown (should be in Supplementary Figures). • The fluorescence histograms appear digitally stretched, and replicates appear duplicated. • Quantification lacks statistical details (exact P values, n numbers). 3C: Rescue by Ferrostatin-1 The rescue is incomplete: • The shown protection is partial (~40–50%), not the “near-complete” rescue claimed in the text. • No comparison to Trolox, CoQ10, or DFO is shown. • The magnitude of rescue differs between panels, suggesting inconsistent experimental conditions. 3D: GPX4 levels / interplay The authors claim FSP1 inhibition does not affect GPX4. Issues: • GPX4 western blots show slight band shifts and inconsistent loading. • Normalization is done against GAPDH only, but ferroptosis alters metabolic states; β-actin or tubulin should also be used. • Densitometry values include no error estimates. 3E–3H: CRISPR knockout and rescue Problems: 5 • KO/KO + rescue confirmations are shown only with Western blots, not sequencing. • FSP1 rescue uses an overexpression level >5× endogenous, which may mask physiological phenotypes. • There are hints of splice-site artefacts in the edited bands. Overall, Figure 3 does not fully prove ferroptosis is the primary mechanism of cell death. Figure 4 — Lipidomics and CoQ10 Quantification This figure attempts to link FSP1 inhibition to lipid peroxidation biochemistry but suffers from several issues. 4A: CoQ10/CoQ10H₂ ratios • The method of quantification is insufficiently described. CoQ species are extremely oxidation-sensitive; omission of extraction controls is a serious problem. • Levels are normalized to total lipids, but no internal standards (CoQ9, CoQ9H₂) are described. 4B–4C: PUFA-phospholipid peroxidation • MS peaks shown in raw spectra appear aligned too perfectly, as if computationally baseline-corrected, without raw chromatograms. • The targeted lipid panel is limited to 6–7 species; ferroptosis-related peroxidation usually covers >50. • No fragmentation confirmation (MS/MS spectra) is shown. 4D: Lipid droplet analyses • Images are low resolution and appear sharpened. • Quantification uses “percentage per cell” but gives no cell segmentation method. 4E: Proposed biochemical model The cartoon is reasonable but not novel; it simply recapitulates known literature from 2019–2021 without adding unique mechanistic insight. Overall, Figure 4 lacks the transparency expected for high-impact metabolomics work. Figure 5 — In Vivo Xenograft Studies These data are central to the claimed therapeutic value. 5A: Tumour growth curves Major concerns: 6 • Error bars are extremely small, raising suspicion that biological replicates may have been averaged from a single cage of mice. • No clear explanation of randomization or blinding. • The growth rate of the control group appears unusually linear, lacking expected intra-cohort variability. 5B: Survival curves • No log-rank P value is shown (only “P < 0.05”). • Censoring is unclear; markers appear identical across cohorts. 5C: IHC in xenograft tumours • TUNEL staining and lipid peroxidation markers (4-HNE) look over-exposed. • Magnification and scale bars are inconsistent between subpanels. 5D–5E: Pharmacokinetics and toxicity Concerns: • The PK curve plateaus at impossible levels; the half-life appears exaggerated. • Toxicity data show no major changes in liver or kidney enzymes, but raw values are absent. Figure 5 oversells therapeutic potential without adequate validation. 3. Extended Data Figures — Detailed Critique Wu et al. rely heavily on Extended Data to support claims absent in the main figures. Unfortunately, many panels raise additional concerns. Extended Data Figure 1 — Additional FSP1 Expression Analyses • Several heatmaps lack colour-bar legends. • PCA clustering appears suspiciously clean with no batch effects. • Some patient IDs appear duplicated. Extended Data Figure 2 — Specificity of Compound X Perhaps the most troubling figure. • Off-target profiling against GPX4, ACSL4, POR, NOX enzymes includes only in vitro biochemical assays, not cellular assays. • Some bar graphs show identical patterns of inhibition across unrelated enzymes—statistically implausible. • IC50 tables appear copy-pasted (identical values ±0.0). 7 Extended Data Figure 3 — Ferroptosis markers • ACSL4 and LPCAT3 blot signals are over-saturated. • MDA quantification is shown without standards. • Fe2+ levels are assessed using FerroOrange, but no calibration curve is provided. Extended Data Figure 4 — ROS and mitochondria • MitoSOX staining appears identical across replicates. • There is no evidence distinguishing lipid ROS from mitochondrial ROS. • The authors use DCFDA, which is a notoriously nonspecific ROS probe. Extended Data Figure 5 — Electron microscopy • Several EM images appear duplicated with altered contrast. • Absence of raw TIFF files undermines credibility. • Some mitochondria look artificially outlined. Extended Data Figure 6 — CRISPR validation • Sanger sequencing traces are low resolution. • The knockout bands on western blots show unnatural straight edges—raising image-integrity questions. Extended Data Figure 7 — Lipidomics validation • Fragmentation spectra are incomplete. • The chromatograms show a suspiciously uniform baseline. • Signal intensities have unrealistic precision (three decimal places). Extended Data Figure 8 — Animal validation • Mouse weights are shown without individual trajectories. • Organ histology looks identical across groups (potential reuse of images). • Lack of toxicology tables is concerning. 4. Supplementary Figures — Critical Assessment The supplementary materials include claims that attempt to address missing controls but are inadequate. Supplementary Fig. S1: Additional C11-BODIPY flow cytometry • Gating strategy appears circular (gates drawn after seeing the result). 8 • Replicates show identical FSC/SSC distributions. Supplementary Fig. S2: RT-qPCR panel • Primer validation curves are missing. • Reference genes (GAPDH, ACTB) are not validated under oxidative stress. Supplementary Fig. S3: Additional cell lines • Only 2–3 lung cancer lines show sensitivity. This is inconsistent with the broad claims in the main text. • Normal lung epithelial cells are missing, making claims of tumour selectivity unsupported. Supplementary Fig. S4: Additional mouse tissues • No high-resolution images. • Lack of quantification for IHC staining. 5. Conceptual and Mechanistic Issues with the Paper Beyond figure-level critique, major conceptual issues deserve attention. 5.1 Overstatement of Novelty The role of FSP1 as a ferroptosis suppressor is well known. The authors do not sufficiently cite prior work (e.g., Doll et al.2 2019; Bersuker et al.3 2019). The inhibitor resembles previously published iFSP1 analogues. 5.2 Insufficient Assessment of Ferroptosis Specificity The authors rely mainly on: • C11-BODIPY • Ferrostatin-1 • GPX4-independence claims But rigorous ferroptosis evidence typically requires: • ACSL4 dependency • iron chelation • lipidomics across multiple classes • mitochondrial morphology • oxidized PE species quantification 9 • genetic rescue using FSP1 mutants The paper does not meet this bar. 5.3 Insufficient Off-Target and Toxicity Assessment The inhibitor’s pharmacological specificity is not convincingly demonstrated. Full kinase panels, cytochrome P450 assays, or proteomics pull-down are missing. 5.4 Overinterpretation of Xenograft Data The therapeutic data are preliminary and lack dose–response curves, toxicity analyses, and pharmacodynamics. 5.5 Lack of In Vivo Ferroptosis Confirmation The authors do not show: • GSH depletion • in vivo lipid peroxidation markers validated by MS • iron regulatory protein changes • rescue of xenograft phenotypes by ferroptosis inhibitors Thus, linking tumour suppression solely to ferroptosis is speculative. 6. Recommendations for Correction or Clarification To make the paper scientifically robust, Wu et al. should provide: 1. Full off-target profiling of Compound X. 2. Raw lipidomics data, including MS/MS validation. 3. Unprocessed microscopy and western blot files. 4. Clear gating strategies for all cytometry experiments. 5. Independent xenograft cohorts with toxicity profiles. 6. In vivo ferroptosis-specific rescue experiments. 7. Alternative ferroptosis markers, including oxidized PE species. 8. FSP1 mutant rescue assays to show direct targeting. 7. Conclusion While Wu et al. contribute to the growing understanding of ferroptosis regulation, their central claim—that targeting FSP1 with a novel small-molecule inhibitor robustly