Biology subjects

Kim, S. M.

Publications and source records attributed to Kim, S. M..

8 recordsLinked to original sources

bioRxiv2026

Physics-Guided Neural Reconstruction of Cellular Membranes for 3D Electron Microscopy

AWith advances in three-dimensional electron microscopy modalities, quantitative characterization of membrane ultrastructure has emerged as an approach to interrogate how organization of proteins and other components around the membrane drive structure and function. Hindering these efforts, the confident reconstruction of geometric features such as membrane curvature is challenging since it requires the calculation of higher-order derivatives from discrete membrane representations. Modern advances in using neural networks to learn continuous implicit representations of complex shapes present a promising solution to this problem. This work presents a physics-informed neural network framework for reconstructing membrane geometries to curvature-order accuracy from images using an implicit neural representation. Benchmarking using synthetic data illustrates that physics-based regularization during training improves accuracy of recovered curvatures, improving robustness to image noise. Application to experimental datasets demonstrate that the framework generalizes to complex cellular structure, such as the Golgi apparatus and mitochondria. We further perform three-dimensional curvature analysis of endocytic pits in cells to reveal anisotropic curvatures at the pit neck, previously predicted to be a lower-energy pathway for neck constriction. This work provides a unified framework for reconstructing three-dimensional membrane shape, including curvature, from volumetric imaging data. By capturing membrane geometry more accurately, our approach yields mechanical insights that can be linked to molecular-scale interactions.

cell biology↗

bioRxiv2026

Long-read sequencing reveals novel mitochondrial genome variants undetected by short-read sequencing in Korean population

While long-read sequencing technologies (e.g., PacBio Revio, ONT) have revolutionized high-quality genome assembly for the human pangenome, mitochondrial genome (mtDNA) analysis still largely relies on short-read and Sanger sequencing. However, short-read sequencing often lacks the resolution required to resolve complex variations due to the unique features of mtDNA, such as high mutation rates and repetitive homopolymeric regions, which frequently lead to alignment artifacts and mapping ambiguities. To address this, we evaluated whether applying long-read sequencing to mtDNA improves analytical quality in empirical data. Through comprehensive bioinformatics analyses, we compared the performance of long-read sequencing against short-read sequencing and microarrays. Our results revealed that long-read sequencing detected the highest number of variants (n = 533), significantly outperforming both short-read sequencing (n = 525) and microarrays (n = 49). Notably, both sequencing methods provided significantly higher resolution in haplogroup assignment compared to microarrays in terms of phylogenetic depth (p < 0.05). Long-read sequencing demonstrated superior detection power, particularly for InDels. We identified two novel non-synonymous variants, including a unique InDel detected exclusively by long-read sequencing. Protein modeling and stability analysis validated that this InDel causes structural instability (RMSD > 2.0 [A],{Delta}{Delta} G = -45.21 kcal/mol). Furthermore, we confirmed that this novel InDel is shared among haplogroup A samples in both the 1000 Genomes Project ONT dataset and the Korean population, highlighting the practical implications of long-read sequencing for molecular biology and population genetics.

genomics↗

bioRxiv2026

Endogenous nucleophilic scavengers of reactive acyl-species neutralize carbon stress

Aberrant protein acylation by reactive acyl species (RAS)--termed carbon stress--is a major driver of aging and metabolic disease. While enzymatic deacylases such as sirtuins counteract aberrant acylation, whether endogenous metabolites can directly neutralize RAS remains unclear. Here, we report that nucleophilic metabolites--taurine, spermidine, and ethanolamine--react with acyl-CoAs, preventing aberrant protein acylation and potentially extending lifespan. We demonstrate that spermidine scavenges acetyl-CoA within the catalytic pocket of p300, a histone acetyltransferase, and extends lifespan in Drosophila. Taurine supplementation in mice fed a high-fat diet promotes N-fatty acyl taurine formation, confirming in vivo scavenging of RAS. These findings identify endogenous nucleophilic metabolites as scavengers that neutralize carbon stress, with implications for combating aging and metabolic disease. SummaryProtein homeostasis is constantly challenged by reactive metabolic intermediates that drive aberrant post-translational modifications, contributing to aging and metabolic disease. While the antioxidant systems that neutralize reactive oxygen species (ROS) are well characterized, whether analogous defenses exist against reactive acyl species (RAS)--the metabolic drivers of non-enzymatic protein acylation, termed carbon stress--has remained unknown. Here we show that three abundant endogenous nucleophilic metabolites--taurine, spermidine, and ethanolamine--directly intercept reactive acyl-CoA species through nucleophilic attack, preventing aberrant protein acylation. We demonstrate that spermidine scavenges acetyl-CoA within the catalytic pocket of the histone acetyltransferase p300, effectively buffering p300-mediated hyperacetylation of muscle, cytosolic, and mitochondrial proteins and extending lifespan in a Drosophila model of p300-driven toxicity. In mice fed a high-fat diet, taurine supplementation promotes the formation of N-fatty acyl taurines, providing direct in vivo evidence for nucleophilic scavenging of diet-derived RAS. Our findings reveal a previously unrecognized metabolite-based defense system--parallel to classical antioxidant defenses against ROS--that neutralizes carbon stress, with broad implications for understanding the molecular mechanisms of aging and the long-observed but mechanistically unexplained health benefits of taurine and spermidine.

biochemistry↗

bioRxiv2025

Multiscale Computational Modeling of the Cardiopulmonary Consequences of Postnatal Hyperoxia with Implications for Preterm Born Children

Moderate to extreme preterm birth (<32 weeks gestation) affects cardiopulmonary structure and function and is associated with increased risk of heart failure through adulthood. The rat hyperoxia (Hx) model (term born; postnatal Hx exposure) captures biventricular changes, including at the cell- and organ-scale, and pulmonary vascular remodeling seen in preterm humans. However, synthesizing these measures across scales and organ systems is challenging. We hypothesized that in-silico modeling of biventricular mitochondrial, myofiber, and organ-scale function plus circulatory function could capture key features of cardiopulmonary abnormalities due to preterm birth. Therefore, we calibrated a multiscale model to subject-specific biventricular pressure-volume data previously obtained from Hx rats alongside normoxic (Nx) controls to investigate the abnormalities in cardiopulmonary function at multiple scales in this animal model of human preterm birth. The calibrated model demonstrates excellent agreement with the data and captures the expected pulmonary vascular changes and right ventricular dilation seen in preterm born children. Our multiscale modeling approach captures cardiopulmonary abnormalities across spatial scales and provides an innovative approach to explore the consequences of preterm birth beyond preclinical experimental data alone. This is a foundational step in understanding the impact of preterm birth on cardiopulmonary disease in childhood as well as adulthood.

bioengineering↗

bioRxiv2025

Targeting FSP1 triggers ferroptosis in lung cancer

Pre-clinical and clinical studies have demonstrated how dietary antioxidants or mutations activating antioxidant metabolism promote cancer, highlighting a central role oxidative stress in tumorigenesis. However, it is unclear if oxidative stress ultimately increases to a point of cell death. Emerging evidence indicates that cancer cells are susceptible to ferroptosis, a form of cell death triggered by uncontrolled lipid peroxidation1-3. Despite broad enthusiasm about harnessing ferroptosis as a novel anti-cancer strategy, whether ferroptosis is a barrier to tumorigenesis and if it can be leveraged therapeutically remains unknown4,5. Using genetically-engineered mouse models (GEMMs) of lung adenocarcinoma (LUAD), we performed tumor specific loss-of-function studies of the two key ferroptosis suppressors, glutathione peroxidase 4 (Gpx4)6,7 and ferroptosis suppressor protein 1 (Fsp1)8,9, and observed increased lipid peroxidation and robust suppression of tumorigenesis, suggesting that lung tumors are highly sensitive to ferroptosis. Furthermore, across multiple pre-clinical models, we found that FSP1 was required for ferroptosis protection in vivo, but not in vitro, underscoring a heightened need to buffer lipid peroxidation under physiological conditions. Lipidomic analyses revealed that Fsp1-knockout (Fsp1KO) tumors had an accumulation of lipid peroxides, and inhibition of ferroptosis with genetic, dietary, or pharmacological approaches effectively restored the growth of Fsp1KO tumors in vivo. Unlike GPX4, FSP1 expression was prognostic for disease progression and poorer survival in LUAD patients, highlighting its potential as a viable therapeutic target. To this end, we demonstrated that pharmacologic inhibition of FSP1 had significant therapeutic benefit in pre-clinical lung cancer models. Our studies highlight the importance of ferroptosis suppression in vivo and pave the way for FSP1 inhibition as a therapeutic strategy in lung cancer patients.

cancer biology↗

bioRxiv2022

Effective Representation Learning via The Integrated Self-Supervised Pre-training models of StyleGAN2-ADA and DINO for Colonoscopy Images

In order to reach better performance in visual representation learning from image or video dataset, huge amount of annotated data are on demand. However, collecting and annotating large-scale datasets are costly and time-consuming tasks. Especially, in a domain like medical, it is hard to access patient images because of the privacy concerns and also not knowing what exactly to annotate without expert effort. One of the solutions to obviate the hassle is to use Self-Supervised learning methods (SSL) and Generative Adversarial Networks (GANs) together. SSL and GANs are quickly advancing fields. GANs have unique capability to create unlabeled data sources containing photo-realistic images while SSL methods are able to learn general image and video features from large-scale data without using any human-annotated labels. In this work, we explore leveraging the power of recently introduced StyleGAN2-ada and Self-Supervised Pre-Training of Dino together for the pretext task. Our underlying insight is also that leveraging the current approaches with Transfer Learning (TF) together brings benefit on doing pretext task in medical domain. By the strategy of unifying these two approaches, we propose the integrated version and use it derive representation learning on polyp dataset.

bioinformatics↗

bioRxiv2019

Checkpoint regulation of nuclear Tos4 defines S phase arrest in fission yeast

From yeast to humans, the cell cycle is tightly controlled by regulatory networks that regulate cell proliferation and can be monitored by dynamic visual markers in living cells. We have observed S phase progression by monitoring nuclear accumulation of the FHA-containing DNA binding protein Tos4, which is expressed in the G1/S phase transition. We use Tos4 localization to distinguish three classes of DNA replication mutants: those that arrest with an apparent 1C DNA content and accumulate Tos4 at the restrictive temperature; those that arrest with an apparent 2C DNA content, that do not accumulate Tos4; and those that proceed into mitosis despite a 1C DNA content, again without Tos4 accumulation. Our data indicate that Tos4 localization in these conditions is responsive to checkpoint kinases, with activation of the Cds1 checkpoint kinase promoting Tos4 retention in the nucleus, and activation of the Chk1 damage checkpoint promoting its turnover. Tos4 localization therefore allows us to monitor checkpoint-dependent activation that responds to replication failure in early versus late S phase.

cell biology↗

bioRxiv2018

Tributyltin induces a transcriptional response without a brite adipocyte signature in adipocyte models

Tributyltin (TBT), a peroxisome proliferator-activated receptor {gamma} (PPAR{gamma})/retinoid X receptor (RXR) ligand and founding member of the environmental obesogen chemical class, induces adipocyte differentiation and suppresses bone formation. A growing number of environmental PPAR{gamma} ligands are being identified. However, the potential for environmental PPAR{gamma} ligands to induce adverse metabolic effects has been questioned because PPAR{gamma} is a therapeutic target in treatment of type II diabetes. We evaluated the molecular consequences of TBT exposure during bone marrow multipotent mesenchymal stromal cell (BM-MSC) differentiation in comparison to rosiglitazone, a therapeutic PPAR{gamma} ligand, and LG100268, a synthetic RXR ligand. Mouse primary BM-MSCs (female, C57BL/6J) undergoing bone differentiation were exposed to maximally efficacious and human relevant concentrations of rosiglitazone (100 nM), LG100268 (100 nM) or TBT (80 nM) for 4 days. Gene expression was assessed using microarrays, and in silico functional annotation was performed using pathway enrichment analysis approaches. Pathways related to osteogenesis were downregulated by all three ligands, while pathways related to adipogenesis were upregulated by rosiglitazone and TBT. However, pathways related to mitochondrial biogenesis and brown-in-white (brite) adipocyte differentiation were more significantly upregulated in rosiglitazone-treated than TBT-treated cells. The lack of induction of genes involved in adipocyte energy dissipation by TBT was confirmed by an independent gene expression analysis in BM-MSCs undergoing adipocyte differentiation and by analysis of a publically available 3T3 L1 data set. Furthermore, rosiglitazone, but not TBT, induced mitochondrial biogenesis. This study is the first to show that an environmental PPAR{gamma} ligand has a limited capacity to induce health promoting activities of PPAR{gamma}.

pharmacology and toxicology↗