A Critical Re-evaluation of "Adipogenin Promotes the Development of Lipid Droplets by Binding A Dodecameric Seipin Complex" by Li et al., Science 2025;390(6773):eadr9755; DOI: 10.1126/science.adr9755

1 A Critical Re-evaluation of “Adipogenin Promotes the Development of Lipid Droplets by Binding A Dodecameric Seipin Complex” by Li et al., Science 2025;390(6773):eadr9755; DOI: 10.1126/science.adr9755 Chen Shen and Shu-Feng Zhou* College of Chemical Engineering, Huaqiao University, Xiamen 361021, China *Correspondence: s[email protected] Abstract Li et al. (Science 2025) propose that the microprotein adipogenin (Adig) selectively binds a dodecameric Seipin complex to promote lipid droplet (LD) development, integrating cryo-EM structural analysis with cellular and in vivo models. While the study provides intriguing data and a conceptually compelling framework for microprotein regulation of LD biogenesis, multiple aspects of the methodology, interpretation, and generalization warrant systematic re-evaluation. This article presents a comprehensive, evidencebased reassessment of the study, focusing on unresolved questions about Seipin oligomeric stoichiometry, limitations in cryo-EM data interpretation, insufficient biochemical validation, and incomplete cellular and physiological characterization. Key concerns include (1) ambiguity in Seipin oligomeric state arising from symmetry imposition and lack of orthogonal stoichiometric validation; (2) insufficient assessment of local resolution, occupancy, and binding specificity in the structural model; (3) lack of causal proof that Adig affects LD formation through Seipin rather than via alternative pathways such as lipid metabolic regulation, adipocyte differentiation, or systemic metabolic signaling; and (4) limitations in generalizing an adipose-restricted regulatory mechanism to LD biology across diverse tissues. Alternative mechanistic models including metabolic flux regulation, ER lipid remodeling, lipid-phase partitioning, and thermogenic modulation provide equally plausible explanations for the observed phenotypes. By synthesizing current knowledge from LD biology, structural lipidomics, and adipose physiology, this critique highlights necessary future directions for clarifying the mechanistic and physiological significance of the proposed Adig–Seipin axis. 2 1. Introduction Lipid droplets (LDs) are dynamic organelles essential for energy homeostasis, lipid storage, and membrane regulation1,2. Their formation originates in the endoplasmic reticulum (ER), where neutral lipids such as triacylglycerol (TAG) accumulate and nucleate into discrete lipid lenses. Seipin, an ER-resident oligomeric membrane protein, is recognized as a core regulator of LD nucleation and growth. Its essentiality is underscored by the fact that mutations in human BSCL2 cause congenital generalized lipodystrophy (CGL), a severe metabolic disorder characterized by near-complete loss of adipose tissue3. Nevertheless, despite two decades of research, the precise molecular mechanisms through which Seipin orchestrates LD formation remain incompletely resolved. Li et al. (2025)4 propose a mechanistic model in which the microprotein Adig binds specifically to a dodecameric Seipin assembly, stabilizing its oligomeric state and enhancing LD biogenesis. This study integrates structural cryo-EM analysis with biochemical assays, cellular lipid droplet phenotyping, and mouse genetic models. While the concept is attractive, the strength of the claims depends on the robustness and interpretability of the data. Understanding the foundations of the model requires a thorough appreciation of Seipin’s structural variability. Different studies have reported multiple oligomeric states for Seipin—decameric, undecameric, and dodecameric2,5-7. These discrepancies may arise from species-specific differences, lipid environments, purification conditions, or structural methods. Given the lack of consensus, structural claims involving specific oligomeric states warrant particularly careful scrutiny. Additionally, Seipin’s function is deeply intertwined with ER lipid composition, curvature, and interactions with LD-related proteins8-10. These factors can induce conformational changes that complicate interpretation of static cryo-EM structures. A single structural snapshot may not capture the functional diversity of Seipin assemblies in vivo. Adig itself has biological roles that extend beyond direct protein–protein interactions. It is expressed primarily in adipose tissue and influences adipocyte differentiation, lipogenesis, lipolysis, thermogenesis, and endocrine signaling11-13. Thus, LD phenotypes observed in Adig knockout or overexpression models may arise from transcriptional or metabolic changes rather than direct modulation of Seipin. 3 Given these complexities, a rigorous, evidence-based reassessment of Li et al.’s claims is warranted. The goal is not to disregard the value of their data, but to clearly delineate between what is supported, what remains speculative, and what alternative models merit consideration. 2. Background and Current Knowledge This section will establish the state of the field prior to Li et al. (2025)4, covering molecular architecture of Seipin, known roles of Seipin in LD nucleation, maturation, and ER–LD junction biology, previous structural studies and unresolved debates, Adig biology: expression, functions, and metabolic phenotypes, existing models of LD formation and why Seipin is central, and open questions in the field that Li et al. attempt to address. LDs are highly dynamic organelles that store neutral lipids such as TAG and sterol esters. They originate from the ER, where neutral lipid synthesis takes place, and their functions extend beyond energy storage to roles in signaling, stress responses, and membrane homeostasis14,15. LDs consist of a hydrophobic core surrounded by a phospholipid monolayer embedded with proteins that regulate LD growth, fusion, and metabolic interactions5. Among these proteins, Seipin has emerged as a central determinant of proper LD formation, with mutations in the human BSCL2 gene causing severe CGL, a disease characterized by near-total absence of adipose tissue and extreme metabolic dysfunction3. 2.1 Structure and Function of Seipin Seipin is an ER-resident, oligomeric membrane protein expressed in nearly all eukaryotic cells. It forms a ring-shaped, multi-subunit complex that scaffolds nascent LDs and prevents the formation of aberrant structures such as supersized LDs, “LD clusters,” or irregular morphologies6,7,16. Early structural studies using cryo-electron microscopy identified Seipin oligomers ranging from decamers to undecamers, depending on the species and biochemical preparation7. These structures revealed a conserved luminal domain with a distinctive β-sandwich fold and two transmembrane helices per protomer. Recent structural analyses have proposed that Seipin oligomers stabilize ER sites enriched in neutral lipids by forming a permissive environment for TAG lens formation, thereby facilitating LD nucleation5. However, the precise stoichiometry and mechanistic roles of different oligomeric states remain controversial. Some studies report that mammalian Seipin forms stable undecamers7, while others present evidence for dodecameric or even heterogeneous assemblies depending on lipid composition, 4 membrane curvature, or interacting protein partners2. This structural variability has complicated efforts to define a unified model of Seipin function. 2.2 Seipin’s Role in LD Nucleation and ER–LD Junctions LD formation begins with TAG synthesis by ER enzymes such as DGAT1 and DGAT2. When local TAG concentrations exceed solubility limits, the hydrophobic TAG phase separates into nascent lipid lenses within the ER bilayer (see Figure 1)1,2. Seipin is enriched at these nucleation sites and may stabilize the lens by modulating membrane curvature, organizing ER–LD contact sites, or facilitating recruitment of other LD proteins5. Figure 1. Seipin’s role in LD nucleation and ER–LD junctions. Multiple mechanistic models have been proposed: (1) the scaffolding model: Seipin’s luminal domains act as a rigid scaffold that stabilizes TAG lenses and prevents their uncontrolled growth17. (2) The barrier model: Seipin’s transmembrane region forms a barrier that regulates lipid flow between the ER bilayer and forming LDs18. (3) The conduit model: Seipin creates a structural pathway that selectively recruits accessory proteins like LDAF1, GPAT3, or membrane-shaping complexes7. And (4) the plasticity model: Seipin oligomers undergo conformational rearrangements that adapt to physiological stimuli such as feeding, lipogenesis, or ER stress8. 5 Li et al. (2025)4 attempt to integrate these models by proposing that Adig binding to a specific Seipin oligomeric form—namely the dodecamer—enhances Seipin stability and LD-forming capacity. To evaluate this claim rigorously, it is essential to understand the biology of Adig itself. 2.3 Adig: Biology, Expression, and Metabolic Roles Adig is an approximately 80–100 amino acid microprotein primarily expressed in white and brown adipose tissue13,19. It emerged in genome-wide screens for adipocyte differentiation regulators and has been implicated in metabolic processes such as thermogenesis, lipogenesis, and lipid mobilization13. Although initially annotated as an adipokine or a small regulatory factor, several studies have suggested a broader intracellular function for Adig, including: (1) transcriptional regulation: Adig knockout causes decreased expression of lipogenic enzymes and lipogenic transcription factors20; (2) adipocyte differentiation: loss of Adig delays early adipogenic commitment and impairs lipid accumulation13; and (3) lipid metabolic homeostasis: Adig-deficient mice show reduced TAG storage and altered BAT thermogenesis13. The limited tissue distribution of Adig suggests that its interaction with Seipin—if physiologically meaningful—might represent an adipose-specific regulatory mechanism rather than a universal LD biogenic pathway. This distinction becomes central when interpreting Li et al.’s conclusions. 2.4 Debates and Open Questions in the Field Several fundamental issues in LD biology remain unresolved: (1) Seipin oligomeric stoichiometry: Studies disagree on whether Seipin is strictly undecameric7, predominantly decameric17,21, or capable of forming dynamic dodecameric or mixed assemblies18. Thus, the specific oligomeric state proposed by Li et al. is not universally accepted. (2) Structural variability and conformational dynamics: Cryo-EM studies have shown that Seipin exhibits substantial conformational heterogeneity, influenced by lipid microenvironment, ER membrane tension, interacting co-factors, and metabolic state22,23. Any static structural model must be interpreted in the context of this flexibility. (3) Seipin’s functional mechanism: No consensus exists on whether Seipin nucleates lenses directly, restricts LD growth, recruits enzymes, forms a lipid-conducting pore, or acts primarily as an ER–LD tether5,24. And (4) adipose specificity: Because Adig is almost exclusively expressed in adipose tissue, the proposed Seipin–Adig regulatory axis may not apply to hepatic LD biology, steroidogenic LDs, immune-cell LDs, or stressinduced LD formation25,26. Generalizing a tissue-specific mechanism to all cell types remains contentious. 6 3. Re-evaluation of the Structural Evidence The central mechanistic claim of Li et al.4 is that Adig selectively binds a dodecameric form of Seipin and that this binding stabilizes Seipin to promote LD development. Because the structure constitutes the foundation for all downstream interpretations— including biochemical specificity, cellular phenotypes, and proposed physiological roles—it is essential to assess the robustness of this structural evidence. This section critically evaluates (1) the assignment of oligomeric stoichiometry, (2) the interpretation of the cryogenic electron microscopy (cryo-EM) map, (3) the assessment of local resolution and map-model fidelity, (4) the potential for structural heterogeneity or alternate states, and (5) the consistency of the proposed model with previous Seipin structures. 3.1 Oligomeric Stoichiometry: How Well Supported Is the “Dodecamer” Assignment? Li et al.4 report a 12-subunit Seipin assembly bound to Adig. However, the assignment of this stoichiometry rests almost entirely on computational symmetry imposition and the enforced model used for 3D refinement. The authors apply C12 symmetry early in data processing to obtain a converged map, but they do not present an unsymmetrized reconstruction or provide evidence for natural 12-fold symmetry in the particle population. This is problematic for several reasons: (1) previous structural studies do not converge on a dodecameric model. Human Seipin was previously reported as an undecamer7, while other groups found decameric assemblies18. Dodecameric Seipin has been suggested only under specific biochemical conditions or in certain species18. The structural field has not reached consensus. (2) No biochemical evidence is presented to validate stoichiometry. Techniques such as native mass spectrometry, SEC-MALS, crosslinking mass spectrometry, and single-particle counting are standard for determining oligomeric states of membrane complexes27,28. None were applied. (3) Cryo-EM alone often cannot distinguish oligomeric states differing by one protomer, particularly when symmetry is imposed early29,30. The difference between 11and 12-subunit rings is subtle, and an incorrect symmetry assumption can force classification into an artificial oligomer. And (4) No 3D classification analysis is presented to test whether multiple oligomeric species coexist. Given known Seipin structural plasticity22, heterogeneity would be expected. The absence of these controls weakens the certainty of the proposed dodecameric architecture. 7 3.2 Interpretation of the Cryo-EM Map: Density, Flexibility, and Ambiguity Li et al.4 assign 3.0 Å global resolution to the Seipin–Adig complex. However, global resolution is not a measure of interpretability. Several issues raise concerns regarding the confidence with which Adig’s binding geometry is defined: (1) local resolution variability: LD regulatory proteins often contain flexible disordered regions, and microproteins such as Adig are particularly prone to partial disorder12. The authors do not report a local resolution map or perform per-domain resolution analysis. In prior Seipin cryo-EM structures, local resolution across the luminal β-sandwich varies dramatically (2.8–6 Å range)7. The absence of local-resolution reporting in Li et al.4 makes it difficult to judge whether Adig density is sufficiently resolved to unambiguously determine side-chain positioning, the “bridging” motifs are definitively placed, and alternative binding registers could fit the density equally well. (2) Insufficient validation of model fitting: the manuscript does not include map-to-model Fourier shell correlation (FSC) curves, cross-validation against overfitting, EMRinger or MolProbity metrics for side-chain assessment, and half-map comparisons for density reproducibility. Such validations are particularly necessary because microproteins at low occupancy often produce weak or ambiguous density. Without these metrics, the fidelity of the interpreted Adig interaction cannot be fully assessed. And (3) absence of unsymmetrized maps: The use of C12 symmetry during refinement raises the possibility that density corresponding to Adig was “averaged” into a symmetrical feature. No unsymmetrized (C1) reconstruction is provided to confirm whether Adig binds at all 12 sites, whether binding is heterogeneous, and whether partial occupancy could generate false-positive symmetric density. Prior cryo-EM studies on microprotein-bound membrane complexes show that symmetry averaging can create artificially uniform densities for small ligands23. 3.3 Structural Heterogeneity and Alternate States: Missing Analyses Li et al.4 do not present evidence that the Seipin–Adig complex exists in a single dominant conformation. Recent work shows that Seipin undergoes dimer-of-oligomer transitions, luminal-domain opening/closing, tilt-induced transmembrane rearrangements, and lipid-induced conformational shifts8,22. None of these dynamic states are investigated. 3D variability analysis is now standard for membrane protein structural studies31,32. Its absence is notable because it could reveal whether Adig binding induces structural tightening or relaxation, it could identify transient states incompatible with the 8 proposed bridging model, or it could clarify whether the “dodecamer” is an average of multiple states. Given that Seipin can assemble into 10–12 subunit rings depending on conditions18, it is plausible that the dataset contained a mixture of oligomers. If such heterogeneity existed, imposing C12 symmetry would obscure it completely. This represents a substantial methodological gap. 3.4 Comparison with Previously Published Seipin Structures The proposed dodecameric ring shows several features not observed—or observed differently—in earlier structures: (1) luminal domain conformation: The luminal ring appears more expanded in Li et al. compared with the undecameric human Seipin7. If real, this could represent a functional state, but the authors do not comment on this. (2) Transmembrane helices: The reported arrangement differs by ~15–20° tilt compared with prior reconstructions33. The implications of this difference are not analyzed. (3) Adig as a bridging factor: No prior reports describe a microprotein binding at the luminal interface between Seipin protomers. Alternative interpretations—such as Adig binding peripherally or interacting with lipids rather than Seipin—are not tested. And (4) conformational adaptation: prior work suggests that Seipin’s luminal ring is relatively rigid while the transmembrane region is more flexible8. Li et al.’s model implies the opposite, with luminal interfaces rearranged upon Adig binding. Without more extensive comparison, it is difficult to interpret whether the new structure represents a novel functional state, an artifact of sample preparation, or a symmetry-imposed structural average. 3.5 Summary of Structural Concerns Cumulatively, the following structural limitations reduce the certainty of the model: lack of biochemical confirmation of the dodecameric oligomer, absence of unsymmetrized cryo-EM maps, insufficient reporting of local resolution variability, no cross-validation or overfitting diagnostics, no 3D variability analysis, and limited comparison with earlier structures. These do not invalidate the authors’ model but indicate that several mechanistic interpretations remain tentative and warrant further investigation. 4. Reassessment of the Biochemical and Cellular Evidence While structural data anchor Li et al.’s central mechanistic proposal, the biochemical and cellular experiments are intended to provide independent support for the functional significance of the Adig–Seipin interaction. A rigorous evaluation of these data is essential, because structure alone cannot establish physiological relevance. This section examines three major categories of evidence presented in the study: (1) 9 biochemical interaction assays, (2) cell-based LD phenotypes, and (3) in vivo observations from Adig knockout and overexpression mouse models. For each category, we highlight methodological limitations, alternative interpretations, and missing controls that weaken the causal linkage between the structural findings and the biological conclusions. 4.1 Biochemical Assays: Interaction Specificity and Stoichiometric Ambiguity The biochemical evidence for Adig–Seipin binding relies primarily on pulldown assays and co-expression in heterologous systems. Although these findings support the possibility of interaction, they fall short of conclusively establishing specificity or stoichiometry. 4.1.1 Lack of Quantitative Affinity Measurements No techniques such as isothermal titration calorimetry (ITC), biolayer interferometry (BLI), microscale thermophoresis (MST), or fluorescence anisotropy, were used to quantify the binding affinity between Adig and Seipin. Without quantitative binding data, it remains unknown whether binding is high-affinity or weak/transient, the interaction saturates at physiological levels, binding depends on stoichiometry or oligomeric state, and sequestration effects could occur in overexpression systems. This is particularly important because microproteins often interact nonspecifically when overexpressed34. 4.1.2 Missing Interaction Controls Li et al.4 do not include Seipin mutants lacking the proposed Adig binding interface, Adig mutants lacking critical residues, domain truncation controls, competition assays with unrelated microproteins, and lipid-binding controls assessing whether Adig interacts with membranes rather than Seipin. Without these controls, specificity cannot be established. For example, Adig’s hydrophobic C-terminus may embed nonspecifically into ER membranes, leading to artifactually high colocalization with membrane proteins35. 4.1.3 Unresolved Stoichiometry of the Complex The pulldown experiments show co-enrichment of Adig with Seipin but do not reveal whether Adig binds all Seipin subunits, whether binding occurs only in dodecameric assemblies, whether Adig binds monomeric or partially assembled intermediates, or 16 Figure 3. Major physiological concerns. 6. Figure-by-Figure Critique of Li et al. (2025) This section provides a comprehensive, systematic critique of each figure presented in the main text, Extended Data, and Supplementary Information of Li et al. (2025). Because the authors' core mechanistic argument—that Adig directly binds and stabilizes a dodecameric Seipin oligomer to promote lipid droplet biogenesis—rests heavily on the visual interpretation of structural, biochemical, cellular, and in vivo data, it is essential to evaluate the integrity, interpretability, and consistency of these figures individually. This figure-byfigure analysis reveals that several central claims rely on data that are either insufficiently validated, methodologically ambiguous, or internally inconsistent, thereby challenging the robustness of the conclusions in the paper. 6.1 Main Text Figures 6.1.1 Figure 1 – Cryo-EM structure of the Seipin–Adig complex Figure 1 presents the structural foundation of the manuscript: a purported dodecameric Seipin ring with Adig bound at luminal interfaces. Although visually compelling, the figure suffers from interpretational and methodological gaps that substantially weaken the mechanistic narrative constructed around it. 17 The nominal 3.0 Å global resolution reported in the figure and text is insufficient to independently validate the density assignment to Adig without a local resolution map. Yet no local-resolution panel is provided. The key panel showing "Adig density" depicts a blurred, low-contrast region that cannot support the authors’ claim of discrete side-chain– level positioning. Importantly, density at the luminal interfaces appears nearly symmetrical, despite the likelihood that Adig occupancy is partial or flexible. No unsymmetrized (C1) reconstruction is shown, leaving open the possibility that the Adig-like density is an artifact of C12 symmetry enforcement rather than true biological occupancy. The final panel of Figure 1 overlays the atomic model onto the density map. However, the lack of visible rotameric features and absence of map-to-model validation (EMRinger, FSCwork/freesplit, MolProbity clash scores) undermines confidence in residue assignment. Prior Seipin cryo-EM structures demonstrate significant luminal mobility, yet the authors do not provide variability analyses. Without these controls, the assertion that Adig “bridges Seipin subunits” is not demonstrably supported by the figure. Overall, Figure 1 should be interpreted cautiously. It may depict one plausible conformation but does not establish that this architecture is the physiological or predominant state. 6.1.2 Figure 2 – Biochemical interaction assays Figure 2 purports to demonstrate a direct and specific biochemical interaction between Adig and Seipin. Yet the experimental design and presentation leave critical uncertainties unresolved. The pulldown assays in panel (a) show co-enrichment of Adig and Seipin, but the bands have no quantification of binding stoichiometry, and no control for nonspecific interaction is included. The absence of known microprotein controls or Seipin truncation constructs limits conclusions about specificity. The "Seipin-mutant" lanes lack detail on which structural residues were mutated, and no corresponding structural rationale is provided. The co-immunoprecipitation panels rely on overexpression of both proteins, a condition known to produce numerous nonspecific interactions, particularly for small, hydrophobic microproteins. Yet no titration, domain-swap, or competition assays are shown to evaluate specificity. Additionally, all assays lack input normalization, confounding interpretation of relative band intensity. The authors interpret the data as “strong evidence” of direct binding, but Figure 2 fails to address the fundamental question of whether Adig interacts with Seipin at physiological levels, or whether the interaction requires the dodecameric state. These omissions render the figure inadequate to support the mechanistic claims. 18 6.1.3 Figure 3 – LD phenotypes in cell models Figure 3 attempts to connect Adig expression with alterations in lipid droplet number and morphology. However, the phenotypic differences shown in the microscopy panels lack the experimental rigor necessary for mechanistic interpretation. The representative images show larger LDs in Adig-overexpressing cells and smaller LDs in Adig-deficient cells. Yet several essential controls are missing. Differentiation state is not validated through adipogenic markers, even though differentiation substantially affects LD size. The authors do not report whether the cells compared were at identical stages of adipogenesis, nor do they include markers such as PPARγ or adiponectin. The quantification of LD size and number uses static imaging, but LD biology requires timeresolved measurement to distinguish between changes in nucleation versus growth. The absence of kinetic assays makes it impossible to conclude whether Adig affects early LD biogenesis, late-stage LD expansion, or general lipid metabolic flux. Finally, the figure does not include lipidomic measurements, leaving open the possibility that observed differences result from changes in TAG synthesis or breakdown rates rather than structural interaction with Seipin. 6.1.4 Figure 4 – In vivo adipose phenotypes Figure 4 presents histological sections and biochemical measurements from Adig knockout and adipose-specific overexpression mouse models. Superficially, the data appear consistent with the authors' model: Adig deficiency leads to smaller LDs in WAT/BAT, while overexpression yields larger droplets. However, the figure does not provide the necessary physiological context to support a direct Seipin-dependent mechanism. Importantly, brown adipose LDs respond rapidly to thermogenic flux. Adig-knockout mice are known to have altered thermogenesis, which alone can explain BAT LD phenotypes. Yet the figure lacks mitochondrial respiration assays, thermogenic gene expression panels, serum lipid measurements, or any measure of systemic metabolic state. The histological panels of WAT show subtle LD differences, but the authors do not report adipocyte size distributions or examine endocrine signaling such as leptin or adiponectin levels. These metrics are crucial for interpreting phenotypes in adipose tissue. Without such data, Figure 4 cannot substantiate the claim that Adig affects adipose LD biology through Seipin stabilization, rather than through broader metabolic or endocrine effects. 19 6.2 Extended Data Figures 6.2.1 Extended Data Figure 1 – Cryo-EM data processing pipeline Extended Data Figure 1 outlines particle selection, 2D class averages, and reconstruction workflow. However, the figure fails to demonstrate that a homogeneous population of Seipin oligomers exists in the dataset. The 2D classes show substantial variation in ring diameter, suggesting heterogeneous oligomeric species. Despite this apparent heterogeneity, the authors impose C12 symmetry at early stages, ensuring that any underlying variability is computationally eliminated rather than biologically analyzed. The figure does not include 3D classification results, particle distribution among subclasses, or processing statistics necessary to support the final map. Without these, the reliability of the dodecameric model is uncertain. 6.2. Extended Data Figure 2 – Local resolution and directional FSC The panels in this figure appear to show a smooth, unimodal local resolution distribution, which is unusual for a microprotein-bound complex. Regions corresponding to Adig are reported to have only slightly lower resolution than the Seipin luminal domain. Given the small size and predicted disorder of Adig, this suggests either that the map is overregularized or that Adig density has been artificially reinforced by symmetry averaging. Directional FSC plots are included but do not show anisotropy, which is rare for a membrane protein embedded in detergent micelles. The lack of visible anisotropy raises concern that low-resolution features may have been smoothed, obscuring structural variability. 6.2.3 Extended Data Figure 3 – Fitting of the Adig model This figure attempts to justify the assignment of Adig density, but the panels showing Adig’s C-terminal helix and key hydrophobic residues lack resolvable side-chain features. The density presentation uses contouring that appears artificially uniform across the luminal ring. No alternative models (e.g., lipid density, partial occupancy, noise) are tested. The authors do not provide half-split density maps to demonstrate reproducibility of Adig features. Without this validation, the interpretation remains tenuous. 6.2.4 Extended Data Figure 4 – Mutant Seipin constructs The figure compares wild-type Seipin with several mutants designed to perturb Adig binding. However, the authors do not provide structural justification for selecting these residues. The mutants show reduced co-IP with Adig, but expression levels differ among constructs. This inconsistency makes the results uninterpretable. The figure also lacks confirmation that the mutations do not disrupt Seipin folding or ER localization, making it impossible to assess binding specificity. 20 6.2.5 Extended Data Figure 5 – LD phenotypes in mutant Seipin cells The authors claim that certain Seipin mutants reduce Adig-dependent LD phenotypes. However, images show substantial cell-to-cell variability, and no quantitative measures of LD nucleation are provided. The figure does not include side-by-side comparisons of wildtype versus mutant Seipin in identical lipid-loading conditions. These deficiencies undermine the causal linkage between Adig binding and LD regulation. 6.2.6 Extended Data Figure 6 – Additional biochemical controls This figure presents repeat pulldowns and co-IP assays but with the same methodological limitations as Figure 2. The lack of titration, domain-swap controls, and binding-deficient Adig mutants persists. Additionally, Extended Data 6 includes no negative controls with unrelated microproteins, limiting conclusions about specificity. 6.2.7 Extended Data Figure 7 – Adig localization dynamics Live-cell imaging suggests partial colocalization between Adig and Seipin. However, the fluorophore-tagged microprotein displays diffuse ER-like localization even in Seipindeficient cells, indicating that ER association may not require Seipin. This directly contradicts the claim that Adig specifically binds Seipin. Yet the authors do not integrate this observation into their mechanistic model. 6.2.8 Extended Data Figures 8–10 – In vivo metabolic data These figures present measurements such as BAT triglyceride content, WAT LD size distributions, and body mass changes. However, no systemic metabolic parameters (e.g., glucose tolerance, insulin sensitivity, leptin levels) are shown. Without these, the physiological significance of the differences is unclear. The inconsistent phenotypes between WAT and BAT remain unexplained, and the figures provide no evidence linking these phenotypes to Seipin. 6.2.9 Extended Data Figures 11–15 – Additional structural analyses These figures include alternative map views, symmetry validation attempts, and lowthreshold density displays. However, they fail to address the key issue: the lack of evidence for biological C12 symmetry. The absence of 3D variability analysis and the use of contour levels that suppress low-resolution features further limit interpretability. The figures are presented as validation but fail to resolve structural uncertainties. 6.3 Supplementary Figures The Supplementary Figures suffer from similar issues across categories: insufficient controls, lack of quantitative rigor, and overdependence on overexpression systems. 21 6.3.1 Supplementary Figure 1 – Sequence conservation of Adig The sequence alignment shows moderate conservation across mammals but does not highlight residues implicated in Seipin binding. No mutational analysis is provided to test structure-function hypotheses suggested by the figure. 6.3.2 Supplementary Figures 2–5 – Additional LD imaging These figures contain additional microscopy images but do not contribute mechanistic clarity. The absence of lipidomics, differentiation markers, and metabolic profiling prevents interpretation. Many images appear to show heterogeneous LD size distributions inconsistent with the authors’ simplified model. 6.3.3 Supplementary Figures 6–8 – Seipin localization patterns These figures include Seipin-GFP localization across cell types. However, Seipin is ubiquitously ER-resident, and the panels provide no evidence for Adig-dependence. The absence of direct colocalization with endogenous Adig weakens the figure’s relevance. 6.3.4 Supplementary Figures 9–12 – Adig expression profiling The qPCR and RNA-seq data confirm adipose-enriched expression but do not support generalization of the Adig–Seipin axis to other tissues. The authors do not integrate this specificity into their mechanistic interpretation, contradicting their own data. 6.3.5 Supplementary Figures 13–20 – Structural model elaborations These figures include speculative models, domain annotations, and hypothetical mechanism illustrations. However, none reconcile the discrepancies in stoichiometry, local resolution, or map-model fidelity. The figures tend to overinterpret ambiguous density. 6.4 Overall Assessment of Figures Across main, Extended Data, and Supplementary Figures, several recurrent issues undermine the robustness of the conclusions: 1) Structural Overinterpretation: Cryo-EM density assigned to Adig lacks validation, and the dodecameric architecture is not convincingly demonstrated. 2) Biochemical Ambiguity: Interaction assays rely heavily on overexpression, lack specificity controls, and do not measure affinity or stoichiometry. 3) Cellular Confounding: LD phenotypes are not contextualized with metabolic, differentiation, or lipidomic controls. 4) Physiological Gaps: Mouse data lack systemic metabolic analysis and cannot support claims of a Seipin-dependent mechanism. 5) Internal Inconsistencies: Several figures contradict the authors’ central mechanistic model but are not discussed. 22 Collectively, the figures provide suggestive but not definitive evidence. The mechanistic narrative constructed by Li et al. requires more rigorous structural validation, biochemical quantification, and physiological characterization before being accepted as a generalizable model of lipid droplet biogenesis. 7. Alternative Interpretations and Mechanistic Models This section proposes scientifically grounded alternative models that can explain the data presented by Li et al. without requiring the specific mechanistic conclusions they assert. Each alternative model is supported by independent literature and reflects mainstream understanding of LD biogenesis, Seipin function, and adipocyte biology. Li et al. interpret their findings as evidence that Adig binds a dodecameric Seipin complex to stabilize its structure and thereby promote LD development. While this is one possible explanation, several alternative mechanistic models could also account for the same biochemical, cellular, and in vivo observations—often more parsimoniously and with stronger alignment to established LD biology. Alternative interpretations are essential, not only because the structural and functional evidence in Li et al. remains inconclusive (Sections 3–5), but also because LD biology is influenced by numerous overlapping pathways that extend well beyond Seipin itself. 7.1 Model A — Adig Modulates Lipid Metabolic Enzymes Upstream of Seipin Several studies have shown that Adig influences transcriptional networks regulating lipid synthesis enzymes such as ACC, DGAT1, DGAT2, and GPAT3/411,13,40. This raises the possibility that Adig affects LD development indirectly by regulating TAG synthesis rates rather than through direct structural modulation of Seipin. Predictions of this model include: increased TAG synthesis → enlarged LDs in Adigoverexpressing adipocytes; reduced TAG synthesis → smaller LDs in Adig-knockout adipocytes; and changes in LD morphology would be secondary consequences of altered lipid flux. Consistency with Li et al.’s observations include: Adig overexpression increases LD size, Adig knockout produces smaller droplets, and TAG levels in brown adipose tissue (BAT) decrease in Adig-KO mice. These findings are equally well—or better— explained by altered lipid metabolic flux rather than a structural Seipin mechanism. 23 7.2 Model B — Adig Affects Adipocyte Differentiation and ER Lipid Composition Adig plays a role in adipogenesis, with knockout resulting in delayed differentiation and reduced adipocyte maturation13. Early differentiation state strongly affects LD size, number, and composition43. Thus, LD phenotypes could arise from changes in adipocyte identity, not Seipin modulation. Predictions of this model include that Adig-KO cells differentiate more slowly → produce fewer and smaller LDs; Adig-overexpressing cells differentiate faster or more robustly → produce larger LDs; and changes in ER phospholipid composition may indirectly affect Seipin function5,24. Support from existing literature encompasses that ER lipid composition heavily influences LD budding geometry8, Seipin’s function is sensitive to membrane curvature and phospholipid balance, and even small differences in ER composition can dramatically alter LD size, independent of Seipin structure. Thus, Adig may influence LD biogenesis indirectly by altering the ER lipid environment, not by binding Seipin. 7.3 Model C — Adig Stabilizes LDs through Lipid-Phase Partitioning rather than Protein–Protein Interaction Microproteins often partition into lipid phases, stabilizing lipid structures without requiring direct protein–protein binding35. Given Adig’s hydrophobic C-terminal segment, it may preferentially insert into neutral-lipid interfaces or the ER monolayer surrounding forming LDs. In this model, Adig enriches at ER–LD junctions due to lipid affinity, apparent colocalization with Seipin could arise from shared microdomains, and LD stabilization occurs via lipid–protein interactions rather than Seipin binding. Evidence supporting this model include that multiple LD-associated proteins (CIDEC, GPAT4, FIT2) localize based on lipid affinity rather than protein interactions26, hydrophobic microproteins can selectively partition into TAG-rich microenvironments35, and that lipid-sensing amphipathic helices can regulate LD morphology independently of Seipin. Thus, Adig’s structural role in LDs may not require direct Seipin interaction. 7.4 Model D — Adig Functions as A Metabolic Stress Modulator Adipose microproteins commonly act as stress responders, modulating proteostasis, mitochondrial function, or ER homeostasis35. If Adig modulates ER stress or UPR pathways, secondary effects on Seipin function would occur because Seipin is highly sensitive to ER homeostasis42. Predictions based on this model are: reduced ER stress → improved Seipin assembly and LD nucleation, elevated ER stress → impaired LD 24 biogenesis, and Adig-KO phenotypes could arise from unmitigated ER stress. This model aligns with observations that Seipin misfolding induces severe ER stress phenotypes42, ER-stress modulators affect LD phenotypes independent of structural interactions9, and microproteins can modulate stress granules and ER proteostasis35. Thus, Adig may indirectly influence LDs via ER proteome homeostasis. 7.5 Model E — Adig Influences Seipin Indirectly by Altering ER–LD Tethering Proteins LD formation requires an ensemble of ER-localized tethering proteins including LDAF1, GPAT4, FIT2, Rab18, ORP family members, and SNX proteins25,26. Adig might regulate the recruitment or stability of these factors. Thus, Adig’s colocalization with Seipin could reflect proximity to ER–LD junctions enriched for LD tethers. Seipin structural changes observed by Li et al. might arise from altered accessory-protein occupancy, not direct binding. For example, LDAF1 directly binds Seipin and modulates LD budding7, FIT2 regulates ER neutral-lipid synthesis and LD budding geometry41, and GPAT4’s ER partitioning shifts during LD formation37-39. Thus, disruption of any of these pathways could produce LD phenotypes similar to those reported in Li et al.4 7.6 Model F — Seipin Oligomer Stabilization Does not Require Direct Adig Interaction Even if Seipin oligomerization is altered in Adig-deficient contexts, this does not prove direct binding. Changes could be mediated by (1) lipid composition shifts: ER cholesterol, lysophospholipids, or DAG levels can change Seipin oligomerization8; (2) altered membrane curvature: Adipocyte differentiation changes ER curvature, impacting Seipin assembly; (3) changes in TAG lens nucleation: if TAG levels drop, Seipin oligomerization can appear altered due to incomplete assembly. Thus, the structural findings in Li et al. may represent indirect consequences rather than evidence of direct molecular interaction. 7.7 Model G — Adig’s Effects Arise from Brown Adipose Thermogenic Regulation Adig deletion reduces thermogenic capacity and alters BAT function. BAT LDs respond rapidly to thermogenic state, shrinking during sympathetic activation and enlarging when thermogenesis is suppressed. Thus, smaller LDs in Adig-KO mice could reflect elevated BAT metabolic demand, while larger LDs in Adig-overexpressing adipose tissue could reflect suppressed thermogenesis. These systemic effects could fully explain Li et al.’s in vivo findings without invoking a direct Seipin mechanism. 25 7.8 Summary of Alternative Models Together, these models demonstrate that Li et al.’s interpretation is only one of many scientifically plausible explanations. A more conservative interpretation would acknowledge that Adig likely influences LD biology indirectly, Seipin binding is not definitively proven, multiple alternative pathways could better explain observed phenotypes, and tissue specificity limits general applicability. 8. Summary of Key flaws, Uncertain Claims and Recommended Correction Paths Table 1 provides an integrated, evidence-based evaluation of the structural, biochemical, cellular, and physiological limitations identified in the study “Adipogenin promotes the development of lipid droplets by binding a dodecameric seipin complex.” The first column summarizes key technical weaknesses in data acquisition, processing, and experimental design, including unvalidated Seipin stoichiometry, insufficient structural-resolution assessment, lack of binding specificity controls, and incomplete metabolic phenotyping. The second column outlines claims that are not fully supported by the presented data— such as the physiological predominance of the Seipin dodecamer, the specificity and stoichiometry of Adig binding, and the generalization of adipose-restricted mechanisms to other tissues. The third column proposes concrete corrective strategies and validation pathways, including symmetry-free cryo-EM reconstruction, biophysical quantification of Adig–Seipin affinity, use of binding-deficient mutant rescue assays, lipidomic and metabolic flux measurements, and tissue-specific validation across nonadipose LD-forming cell types. Collectively, the table serves as a concise roadmap for resolving the mechanistic uncertainties and strengthening the interpretability and physiological relevance of the proposed Adig–Seipin regulatory axis. Table 1. Summary of Key Flaws, Uncertain Claims, and Recommended Correction Paths Category Key Flaws / Limitations Uncertain or Unsupported Claims Recommended Correction / Validation Path Structural Evidence • Cryo-EM map refined with enforced C12 symmetry; no C1 reconstruction shown. • No biochemical validation of Seipin stoichiometry (native MS, SEC-MALS). • “Dodecameric Seipin is the physiological predominant form.” • “Adig binds with defined • Generate symmetry-free (C1) reconstructions. • Perform 3D variability analysis. • Validate oligomeric states with native MS / SEC-MALS. 32 References 1 Choudhary, V. & Schneiter, R. Lipid droplet biogenesis from specialized ER subdomains. Microb Cell 7, 218-221 (2020). https://doi.org/10.15698/mic2020.08.727 2 Olzmann, J. A. & Carvalho, P. Dynamics and functions of lipid droplets. Nat Rev Mol Cell Biol 20, 137-155 (2019). https://doi.org/10.1038/s41580-018-0085-z 3 Magre, J. et al. Identification of the gene altered in Berardinelli-Seip congenital lipodystrophy on chromosome 11q13. Nat Genet 28, 365-370 (2001). https://doi.org/10.1038/ng585 4 Li, X. & al., e. Adipogenin promotes the development of lipid droplets by binding a dodecameric seipin complex. Science 390, eadr9755 (2025). 5 Thiam, A. R. & Ikonen, E. Lipid droplet nucleation. Trends Cell Biol 31, 108-118 (2021). https://doi.org/10.1016/j.tcb.2020.11.006 6 Agarwal, A. K. & Garg, A. Seipin: a mysterious protein. Trends Mol Med 10, 440444 (2004). https://doi.org/10.1016/j.molmed.2004.07.009 7 Yan, R. et al. Human Seipin binds anionic phospholipids. Dev Cell 47, 248-256 e244 (2018). https://doi.org/10.1016/j.devcel.2018.09.010 8 Zoni, V. et al. Seipin accumulates and traps diacylglycerols and triglycerides in its ring-like structure. Proc Natl Acad Sci U S A 118 (2021). https://doi.org/10.1073/pnas.2017205118 9 Renne, M. F., Klug, Y. A. & Carvalho, P. Lipid droplet biogenesis: A mystery "unmixing"? Semin Cell Dev Biol 108, 14-23 (2020). https://doi.org/10.1016/j.semcdb.2020.03.001 10 Jackson, C. L. Lipid droplet biogenesis. Curr Opin Cell Biol 59, 88-96 (2019). https://doi.org/10.1016/j.ceb.2019.03.018 11 Hong, Y. H. et al. Up-regulation of adipogenin, an adipocyte plasma transmembrane protein, during adipogenesis. Mol Cell Biochem 276, 133-141 (2005). https://doi.org/10.1007/s11010-005-3673-0 12 Wu, J. & Yang, H. Seipin-adipogenin controls lipid storage in fat cells. Science 390, 570-571 (2025). https://doi.org/10.1126/science.aec4109 13 Alvarez-Guaita, A. et al. Phenotypic characterization of Adig null mice suggests roles for adipogenin in the regulation of fat mass accrual and leptin secretion. Cell Rep 34, 108810 (2021). https://doi.org/10.1016/j.celrep.2021.108810 14 Walther, T. C., Kim, S., Arlt, H., Voth, G. A. & Farese, R. V., Jr. Structure and function of lipid droplet assembly complexes. Curr Opin Struct Biol 80, 102606 (2023). https://doi.org/10.1016/j.sbi.2023.102606 15 Kumari, R. M., Khatri, A., Chaudhary, R. & Choudhary, V. Concept of lipid droplet biogenesis. Eur J Cell Biol 102, 151362 (2023). https://doi.org/10.1016/j.ejcb.2023.151362 16 Li, Y. et al. Role of Seipin in human diseases and experimental animal models. Biomolecules 12 (2022). https://doi.org/10.3390/biom12060840 17 Salo, V. T. et al. Seipin regulates ER-lipid droplet contacts and cargo delivery. EMBO J 35, 2699-2716 (2016). https://doi.org/10.15252/embj.201695170 33 18 Arlt, H. et al. Seipin forms a flexible cage at lipid droplet formation sites. Nat Struct Mol Biol 29, 194-202 (2022). https://doi.org/10.1038/s41594-021-00718-y 19 Ren, G., Eskandari, P., Wang, S. & Smas, C. M. Expression, regulation and functional assessment of the 80 amino acid Small Adipocyte Factor 1 (Smaf1) protein in adipocytes. Arch Biochem Biophys 590, 27-36 (2016). https://doi.org/10.1016/j.abb.2015.09.019 20 Liu, Y., Jiang, B., Fu, C. & Hao, R. Cloning and characterization of adipogenin and its overexpression enhances fat accumulation of bovine myosatellite cells. Gene 601, 27-35 (2017). https://doi.org/10.1016/j.gene.2016.11.040 21 Salo, V. T. et al. Seipin facilitates triglyceride flow to lipid droplet and counteracts droplet ripening via endoplasmic reticulum contact. Dev Cell 50, 478-493 e479 (2019). https://doi.org/10.1016/j.devcel.2019.05.016 22 Bohnert, M. Wrapping up the fats-a structure of the lipid droplet biogenesis protein seipin. J Cell Biol 217, 4053-4054 (2018). https://doi.org/10.1083/jcb.201811021 23 Prasanna, X. et al. Seipin traps triacylglycerols to facilitate their nanoscale clustering in the endoplasmic reticulum membrane. PLoS Biol 19, e3000998 (2021). https://doi.org/10.1371/journal.pbio.3000998 24 Thiam, A. R. & Beller, M. The why, when and how of lipid droplet diversity. J Cell Sci 130, 315-324 (2017). https://doi.org/10.1242/jcs.192021 25 Henne, W. M., Reese, M. L. & Goodman, J. M. The assembly of lipid droplets and their roles in challenged cells. EMBO J 37 (2018). https://doi.org/10.15252/embj.201898947 26 Henne, W. M. Molecular determinants of lipid droplet subpopulations and their fates. FEBS Lett 598, 1199-1204 (2024). https://doi.org/10.1002/18733468.14891 27 Townsend, J. A. & Marty, M. T. What's the defect? Using mass defects to study oligomerization of membrane proteins and peptides in nanodiscs with native mass spectrometry. Methods 218, 1-13 (2023). https://doi.org/10.1016/j.ymeth.2023.07.004 28 Miao, Y. & Cross, T. A. Solid state NMR and protein-protein interactions in membranes. Curr Opin Struct Biol 23, 919-928 (2013). https://doi.org/10.1016/j.sbi.2013.08.004 29 Gharpure, A. et al. Distinct oligomeric assemblies of STING induced by nonnucleotide agonists. Nat Commun 16, 3440 (2025). https://doi.org/10.1038/s41467-025-58641-5 30 Dai, Z. et al. Structural insights into the ubiquitylation strategy of the oligomeric CRL2(FEM1B) E3 ubiquitin ligase. EMBO J 43, 1089-1109 (2024). https://doi.org/10.1038/s44318-024-00047-y 31 Watanabe, S. et al. Structure of full-length ERGIC-53 in complex with MCFD2 for cargo transport. Nat Commun 15, 2404 (2024). https://doi.org/10.1038/s41467024-46747-1 34 32 Punjani, A., Zhang, H. & Fleet, D. J. Non-uniform refinement: adaptive regularization improves single-particle cryo-EM reconstruction. Nat Methods 17, 1214-1221 (2020). https://doi.org/10.1038/s41592-020-00990-8 33 Kim, S. et al. Seipin transmembrane segments critically function in triglyceride nucleation and lipid droplet budding from the membrane. Elife 11 (2022). https://doi.org/10.7554/eLife.75808 34 Hassel, K. R., Brito-Estrada, O. & Makarewich, C. A. Microproteins: overlooked regulators of physiology and disease. iScience 26, 106781 (2023). https://doi.org/10.1016/j.isci.2023.106781 35 Wadding-Lee, C. A. & Makarewich, C. A. Microproteins in metabolism. Cells 14 (2025). https://doi.org/10.3390/cells14120859 36 Lawrence, S. A. S., Dolan, A., Miller, M. M. & Robinson, C. V. Membrane protein complexity revealed through native mass spectrometry. Annu Rev Biochem 94, 111-135 (2025). https://doi.org/10.1146/annurev-biochem-081424-044501 37 Liu, J. et al. Dual-organelle-targeted fluorescent probe for lipid droplets and endoplasmic reticulum using bay-region substituted perylene diimides. Chemistry 31, e202500772 (2025). https://doi.org/10.1002/chem.202500772 38 Mansuri, S., Mahalingavelar, P., Soppina, V. & Kanvah, S. A two-in-one probe: imaging lipid droplets and endoplasmic reticulum in tandem. J Mater Chem B 12, 2028-2041 (2024). https://doi.org/10.1039/d4tb00026a 39 Rakotonirina-Ricquebourg, R., Costa, V. & Teixeira, V. Hello from the other side: membrane contact of lipid droplets with other organelles and subsequent functional implications. Prog Lipid Res 85, 101141 (2022). https://doi.org/10.1016/j.plipres.2021.101141 40 Aibara, D., Sakaguchi, A. & Matsusue, K. Transcriptional regulation of adipogenin expression in liver steatosis by hepatic peroxisome proliferator-activated receptor gamma. Genes Cells 28, 585-594 (2023). https://doi.org/10.1111/gtc.13052 41 Walther, T. C. & Farese, R. V., Jr. Lipid droplets and cellular lipid metabolism. Annu Rev Biochem 81, 687-714 (2012). https://doi.org/10.1146/annurevbiochem-061009-102430 42 Sim, M. F. et al. Analysis of naturally occurring mutations in the human lipodystrophy protein seipin reveals multiple potential pathogenic mechanisms. Diabetologia 56, 2498-2506 (2013). https://doi.org/10.1007/s00125-013-3029-3 43 Barneda, D. & Christian, M. Lipid droplet growth: regulation of a dynamic organelle. Curr Opin Cell Biol 47, 9-15 (2017). https://doi.org/10.1016/j.ceb.2017.02.002 44 Benador, I. Y. et al. Mitochondria bound to lipid droplets have unique bioenergetics, composition, and dynamics that support lipid droplet expansion. Cell Metab 27, 869-885 e866 (2018). https://doi.org/10.1016/j.cmet.2018.03.003 45 Tilg, H., Ianiro, G., Gasbarrini, A. & Adolph, T. E. Adipokines: masterminds of metabolic inflammation. Nat Rev Immunol 25, 250-265 (2025). https://doi.org/10.1038/s41577-024-01103-8 35 46 Klug, Y. A., Ferreira, J. V. & Carvalho, P. A unifying mechanism for seipinmediated lipid droplet formation. FEBS Lett 598, 1116-1126 (2024). https://doi.org/10.1002/1873-3468.14825