Nutrient State, Aging, and Diet Modulate SAM50-Dependent Mitochondrial Remodeling and Systemic Metabolic Signatures
Antentor Othrell Hinton, Sepiso K Masenga, Victoria Baskerville, Mark Petrovic, Benjamin Rodriguez, David L Hubert, Tyne W Miller-Fleming, Young Do Koo, Prasanna Katti, Prasanna Venkhatesh, Annet Kirabo, Edgar Garza Lopez, Amber Crabtree, Andrea Marshall, Campbell Blake
TL;DR
This study explores how SAM50, a protein involved in mitochondrial function, affects metabolic health in relation to diet, aging, and genetic variation.
Contribution
The study reveals how SAM50 genetic variation and expression influence mitochondrial remodeling and systemic metabolic health in response to diet and aging.
Findings
Common genetic variation in SAMM50 is linked to liver-related metabolic disorders.
Aging and high-fat diets reduce Sam50 expression and disrupt mitochondrial structure.
Nutrient state and aging signals influence SAM50-dependent mitochondrial remodeling.
Abstract
Although Sorting and Assembly Machinery 50 (SAM50) is known to regulate nutritional and metabolic stress related to ageing, its exact role is not well understood. This experimental study combines both human and animal models to understand the role that SAM50 plays in nutrient, age-related metabolic remodeling. We also wanted to define the clinical relevance of SAMM50 genetic variation in human disease. Our study integrated clinical and genetic data from three large and independent human biobanks to assess the clinical implications of genetic variation in SAMM50. We then conducted mechanistic studies in mice using Serial Block-Face Scanning Electron Microscopy and Transmission Electron Microscopy for three-dimension analysis of mitochondrial morphology, immunoblotting, metabolomics/lipidomics, and assessment of metabolic parameters in models of fasting, aging, and a high-fat diet (HFD).…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsMitochondrial Function and Pathology · Genetics, Aging, and Longevity in Model Organisms · GDF15 and Related Biomarkers
Introduction
Mitochondria and the proteins regulating its structure and function, are important for cellular energy homeostasis and overall metabolic health^1^. The structure and function of mitochondria are largely determined by ongoing fusion and fission processes, as well as the complex organization of their inner membrane cristae. These activities, are closely linked to cellular nutrient levels, stress responses, and aging^2,3^. The mitochondrial sorting and assembly machinery (SAM) is essential for maintaining the structure of the inner membrane and forming important contact sites with other cell organelles^4^. SAM50, a key and highly conserved component of this SAM complex, has two main roles. SAM50 helps assemble beta-barrel proteins into the outer mitochondrial membrane and additionally, it interacts directly with the MICOS complex^5–7^. This positions SAM50 as a key regulator of mitochondria structure and function.
Prior studies have demonstrated that any stress-related changes in mitochondrial architecture including aging has been linked to metabolic diseases and liver-related disorders such as insulin resistance and hepatic steatosis, respectively^8–10^. Although mitochondrial dynamics have been recognized as important in metabolic diseases, the specific role of SAM50 in regulating systemic metabolism in response to environmental and genetic variations remains unclear. We therefore wanted to know if SAM50 expression and function respond to nutritional interventions, including whether diet-induced mitochondrial defects can be reversed. We also sought to elucidate the role of aging on SAM50 expression. This study combines human genetics with mechanistic studies in mice and primary cells to examine how SAM50 functions as a regulator of mitochondrial remodeling responsive to nutrients and aging.
Methods
Biobank Genetic Association Studies
Vanderbilt University Medical Center curates an electronic health record (EHR)-linked biobank, BioVU. This opt-in program collects and stores blood samples from routine clinical visits. Currently, BioVU houses more than 350,000 biological samples linked to de-identified EHRs dating back several decades^11,12^. Genetic data for 90,000 individuals was analyzed and cleaned as previously described^13^. In summary, individuals were genotyped on Illumina’s Multi-Ethnic Genotyping Array (MEGA) and imputed on the Michigan Imputation Server using the Haplotype Reference Consortium (HRC) reference panel. Quality control was performed to filter the genetic data based on previously reported minor allele frequency, individual and variant level missingness, and deviations within Hardy-Weinberg equilibrium. Genetically related individuals were removed from the analysis. Principal component analysis was performed with the 1000 genomes reference populations and downstream analyses were performed within two BioVU populations: individuals of European genetic ancestry (n=70,404) and individuals of African genetic ancestry (n=15,175). Genetically-regulated gene expression (GreX) for SAMM50 was calculated using predictive models PrediXcan, UTMOST, JTI using GTEx version 8 data^14–17^. The model with the highest performance r2 was used for each SAMM50-tissue pair. MultiXcan was then employed to synthesize a cross-tissue SAMM50 GReX model^14,18^. Within genetically-defined ancestry groups, associations between SAMM50 GReX and 1,704 clinical phenotypes (via logistic regression) and 326 laboratory traits (via linear regression) were tested, adjusting for principal components (1–10), sex, age, median age of medical record, and genotyping batch. Clinical laboratory values were extracted from the EHR as previously described (REF). Phenotypes were extracted from the EHR by mapping structured clinical data (ICD9 and ICD10 codes) to phecodes as previously described^13,19^.
To validate the clinical relevance of SAMM50 in human disease from additional biobank populations, we explored publicly available summary-level genome-wide association results from two large, independent biobanks (Biobank Japan and UK Biobank). BioBank Japan is a biobank with 260,000 participants from 12 medical institutions across Japan. BioBank Japan is managed by the Institute of Medical Science at the University of Tokyo and the RIKEN Center for Integrative Medical Sciences, the University of Tokyo, and Osaka University performed genotyping on the BioBank Japan samples. The summary results of several BioBank Japan genome-wide association studies (GWAS) are publicly available at the BioBank Japan PheWeb portal (https://pheweb.jp/)^20,21^. Additional information regarding the pheweb repository is available on the github website (https://github.com/statgen/pheweb/). For additional support of our SAMM50 genetic associations, we explored the publicly available GWAS results from UK Biobank data. Similar to the BioBank Japan pheweb portal, the pheweb for UK Biobank includes summary information for genetic analyses performed across 400,000 individuals from the UK. These results include GWAS summary statistics from association studies performed on 1,419 phenotypes extracted from the EHR. We used the UK Biobank data that imputed with the HRC reference panel, as this was the same reference panel used for the BioVU studies. The UK Biobank pheweb was built using PheWeb version 1.3.15^22^. The rationale behind interrogation of these three independent biobanks was to establish robust, convergent and orthogonal evidence supporting associations between SAMM50 and liver-related metabolic disorders and outcomes.
Animal care and fasting and refeeding paradigm
We conducted all animal experiments at the University of California, Los Angeles (UCLA) and all procedures were performed in accordance with the approved protocols established by the UCLA Institutional Animal Care and Use Committee (IACUC) and in compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Ethical approval was obtained prior to the start of the study. We housed the adult mice in a controlled vivarium with a 12-hour light/dark cycle and provided them with free access to food and water unless specified otherwise. Temperature was also controlled and checked daily. For the fasting experiments, fasting was initiated at the beginning of the light cycle to minimize circadian variability and mice were singly housed and subjected to a 24 h fast with continuous access to water. The control mice had unrestricted access to food throughout the study. For refeeding experiments, mice were fasted for 24 h and subsequently refed standard chow for 4 h prior to tissue collection. All animals were euthanized at the same circadian time point to control for diurnal effects on metabolism and mitochondrial protein expression.
Tissue collection
After the fasting/refeeding period was completed, mice were euthanized for tissue processing. Briefly, tissue was dissected immediately after euthanization, and then briefly rinsed in ice-cold phosphate-buffered saline. This was done in order to remove any residual blood. The tissue was snap frozen in liquid nitrogen and stored at −80°C until further processing. All procedures were conducted according to UCLA IACUC-approved protocols.
Protein isolation
To initiate protein isolation, the frozen tissues were first homogenized on ice in radioimmunoprecipitation assay buffer which was supplemented with protease and phosphatase inhibitor cocktails. Next, the homogenates were incubated on ice with intermittent mixing followed by clarification through by centrifugation at 14,000 × g for 15 min at 4°C. The resulting supernatants after centrifugation contained total protein lysates which were collected. Protein concentrations were then measured using a bicinchoninic acid assay, following the manufacturer’s instructions.
SDS–PAGE and immunoblot analysis
We used Sodium Dodecyl Sulfate–Polyacrylamide Gel Electrophoresis to equalize amounts of protein using gradient polyacrylamide gels and then we transferred the protein onto polyvinylidene difluoride membranes. We used 5% nonfat dry milk prepared in Tris-buffered saline containing 0.1% Tween-20 to block the membranes and then incubated them overnight at 4°C. We incubated with primary antibodies against mitochondrial respiratory chain component proteins complexes I, II, III, and V, as well as SAM50 and Tom20. After incubation with horseradish peroxidase-conjugated secondary antibodies specific to the species, the immunoreactive bands were visualized using enhanced chemiluminescence. GAPDH served as the loading control for this procedure.
Molecular Biology
RNA extraction was performed using TRIzol, followed by reverse transcription and quantitative PCR (qPCR) with SYBR Green and primers specific for Sam50/SAMM50 and GADPH^23^. For western blot analyses, tissue lysates were prepared in RIPA buffer, resolved by Tris-glycine gels, transferred to nitrocellulose membranes, and probed with antibodies against SAM50 (Proteintech 20824–1-AP), respiratory complex subunits, Tom20, and Tubulin. Detection was carried out using IRDye-conjugated secondary antibodies and visualized on a Li-Cor Odyssey CLx system.
Data were presented as fold changes normalization to GAPDH. The qPCR primers utilized were derived from previously published sequences^24^, as detailed in Table 1.
Cell culture and serum fasting, and refeeding
We used an epithelial hepatoma cell line derived from liver tissue, Hepa-1c1c7 cells, and placed them in a humidified incubator with 5% CO_2_ in Dulbecco’s modified Eagle’s medium that was supplemented with 1% L-glutamine, 10% fetal bovine serum and 1% penicillin–streptomycin, at 37°C. The Cells were routinely passaged and used for experiments at 70–80% confluence.
For the serum fasting/refeeding experiments, we first washed cells twice with phosphate-buffered saline and incubated them in serum-free DMEM for 6 hours. The control cells were maintained in complete medium having 10% fetal bovine serum. For the refeeding procedure, we first starved the cells for 6 hours and then immediately after refed them with complete medium containing 10% fetal bovine serum for 4 hours. We conducted all experimental procedures and conditions in parallel and collected data at identical time points to minimize variability.
Microscopy and Image Analysis
Transmission Electron Microscopy (TEM)
Tissue fixation, embedding, and sectioning were performed according to standardized protocols^25^ to minimize bias. The cells were fixed directly in culture dishes. Fixation reagents included using 2.5% glutaraldehyde in 0.1 M cacodylate buffer, followed by post-fixation in 1% osmium tetroxide. To dehydrate the samples, we used a graded ethanol series followed by embedding them in epoxy resin. Ultrathin sections (90–100 nm) were stained and imaged. Stained was done using uranyl acetate and lead citrate. Images were acquired using a transmission electron microscope at identical magnifications for all conditions. This was done utilizing NIH’s ImageJ software^26,27^. Mitochondrial morphology, area, and number were quantified from randomly selected fields using blinded analysis. Mitochondria were manually outlined to determine both their area per cell and the number of mitochondria relative to cytoplasmic area. These measurements were taken from several cells under each condition, using independent biological replicates. Results are shown as mean ± SEM.
Serial Block-Face Scanning Electron Microscopy (SBF-SEM):
Three-dimensional ultrastructural analysis was performed as previously described^25,27,28^. Small liver tissue blocks were fixed in glutaraldehyde, and then subsequently fixed with osmium tetroxide. They were then stained, and dehydrated through a graded ethanol series and embedded in epoxy resin, and serially sectioned (50 nm) for imaging using a VolumeScope 2 serial block-face scanning electron microscope equipped with an in-chamber ultramicrotome. Serial images were taken at a voxel resolution suitable for mitochondrial reconstruction, with section thickness kept constant across samples.
For 3D reconstruction, 300–400 slices per sample were manually segmented by blinded investigators in Amira software. Morphometric parameters (volume, surface area, sphericity, complexity index) were calculated for approximately 250 mitochondria per experimental condition.
Mitochondrial DNA content
Real-time polymerase chain reaction was employed to quantify Mitochondrial DNA content from liver tissue. We used the DNeasy Kit according to the manufacturer’s instructions for the extraction and purifying of total DNA from liver samples. In order to amplify mitochondrial and nuclear DNA markers, we used five nanograms of total DNA then employed quantitative PCR from the ABI Prism 7900HT instrument in 384-well plate format with SYBR Green I chemistry and ROX as an internal reference dye. Using the SDS 2.1 software in combination with scripted workflows implemented in Microsoft Access and Microsoft Excel, data acquisition and analysis were automated increasing robustness, accuracy and reproducibility. β-actin served as the reference for nuclear DNA, while cytochrome c oxidase subunit I was used as the marker to measure mitochondrial DNA content. Mitochondrial DNA abundance was expressed relative to the nuclear Rpl13a gene. The primer sequences that were used for amplification in this protocol were as shown below:
Cox1 forward, GCC CCC AGA TAT AGC ATT CCC
Cox1 reverse, GTT CAT CCT GTT CCT GCT CC
Rpl13a forward, GAG GCC CCT ACC ATT TCC GA
Rpl13a reverse, GGC TTC AGC CGA ACA ACC TT.
Serial block face scanning electron microscopy analysis of hepatic lipid droplets
As described above, liver tissue was prepared as per standard protocols for serial block face scanning electron microscopy (SBF SEM). Standard heavy metal staining and resin embedding protocols optimized for ultrastructural preservation of lipid rich organelles were employed. For the acquisition of datasets, we generated volumetric image stacks spanning approximately 10 × 10 × 50 μm of hepatic parenchyma that were acquired at nanometer scale resolution. Following this, individual lipid droplets were then manually segmented across consecutive serial sections using Amira software as earlier described. We calculated lipid droplet volume, surface area, and spatial distribution lipid droplet from reconstructed objects. Quantification of lipid droplet contact site coverage with mitochondria was conducted by assessing the area of direct apposition between lipid droplets and mitochondria relative to the total lipid droplet surface area. We defined Lipid droplet coverage as the proportion of imaged cellular volume occupied by lipid droplets. To minimize bias, all analyses were performed by a trained researcher who was blinded to the identity and knowledge of the age groups.
Biochemical Assays
Thiobarbituric acid reactive substances analysis
Lipid peroxidation was quantified through malondialdehyde determination using the thiobarbituric acid reactive substances (TBARS) assay. For cellular analyses, cells were harvested and prepared according to the manufacturer’s protocol with the TBARS assay kit from Cayman Chemical Company. Tissue measurements involved homogenizing liver samples in an appropriate assay buffer, followed by clarification via centrifugation prior to analysis. TBARS results were normalized to protein content for cellular samples and to tissue weight for liver samples. All assays were conducted in technical duplicates and validated across independent biological replicates.
Hepatic and serum triglyceride measurements
Hepatic and serum triglycerides were measured using a commercial enzymatic kit (EnzyChrom^™^, BioAssay Systems) as previously described^24,29^. Briefly, triglyceride levels were measured in liver tissue and serum after a six hour fast. To extract Liver triglycerides, we used a solution of isopropanol and Triton X 100. The resulting extracts clarified by centrifugation were analyzed according to the manufacturer’s protocol.
To quantify Serum triglycerides, we measured them directly without extraction. To calculate and quantify Triglyceride concentrations, we used a standard curve generated with known triglyceride standards and normalized to tissue weight for liver samples.
Total bile acids in liver tissue were measured using a colorimetric assay (Crystal Chem). The Liver samples first underwent sequential extraction with 95% ethanol overnight, following this, 80% ethanol was then employed for two hours, followed by methanol-chloroform (2:1) for two hours at 50°C. Quantification was performed using a Genzyme Diagnostics bile acid assay kit following the manufacturer instructions, and the final results were then normalized to tissue mass.
Metabolomics and Lipidomics
For untargeted metabolomics, frozen tissues were extracted using a 40:40:20 acetonitrile:methanol:water mixture containing 0.5% formic acid and 15% ammonium bicarbonate^30^. Extracts were analyzed via HILIC chromatography coupled to an Orbitrap Exploris 240 mass spectrometer in both positive and negative ion modes, with data processing performed using EL-MAVEN (Thermo Scientific)^31^. For lipidomics, tissues were homogenized and lipids extracted with a mixture of Isopropanol/H_2_O/Ethyl acetate (30:10:60), spiked with Avanti Lipidomix internal standards. Dried extracts were reconstituted and analyzed using RP-UHPLC on a CSH C18 column interfaced with the same Orbitrap MS in AcquireX mode for MS/MS.
Statistical Analysis
All statistical analyses for biological assays were performed in GraphPad Prism 10.2.3. For immunoblots, densitometry software with normalization to GAPDH was employed to quantify band intensities. Data are presented as mean ± SEM from independent biological replicates. Statistical comparisons between fasting and refed groups were performed using unpaired two-tailed Student’s t-tests, with statistical significance defined as p < 0.05. Regression models for the PheWAS and LabWAS of BioVU data was performed in R (version 3.6). Bonferroni-corrected p-values were used to determine statistical significance for the PheWAS and LabWAS results and were calculated by dividing 0.05 by the number of phenotypes or lab values tested, respectively.
Data from SBF-SEM/TEM studies represent a minimum of three independent biological replicates with analyses executed by blinded investigators. Data are presented as mean ± SEM. Statistical comparisons between age groups were performed using unpaired two-tailed Student’s t-tests, Comparisons among multiple groups utilized one-way ANOVA followed by Fisher’s protected LSD test. Statistical significance was denoted as p < 0.05, p < 0.01, p < 0.001, and p < 0.0001.
RESULTS
SAMM50 genetically regulated expression is associated with liver disease susceptibility in humans.
To examine the clinical relevance of SAMM50 genetic variation, we interrogated genetic association data from three large and independent biobanks. Within a subset of BioVU, a biobank curated by Vanderbilt University Medical Center (Figure 1, Supplemental Table 1), participants had a mean current age of 55.2 years and a median age of medical record of 46.0 years, with an average record length of 8.5 years, providing substantial longitudinal data for phenotype analysis. The cohort consisted of 78% individuals of European genetic ancestry and 17% individuals of African genetic ancestry, Supplemental Table 1. In individuals of European genetic ancestry (n=70,404), the PheWAS for SAMM50 GReX demonstrated 10 significant phenotype associations, most of which were liver-related disease phenotypes (Figure 2A and Supplemental File 1), Significant associations with SAMM50 GReX included chronic liver disease and cirrhosis, cirrhosis of liver without mention of alcohol, other chronic nonalcoholic liver disease, liver abscess and sequelae of chronic liver disease, liver replaced by transplant and alcoholic liver damage. Associated sequelae of advanced liver disease, such as portal hypertension acute gastritis, esophageal bleeding/varices, and disorders of the orbit, were also significantly associated with SAMM50 GReX. The strong, concordant signal across multiple liver-specific phenotypes underscores the hepatic relevance of SAMM50 genetic regulation in human populations. In contrast, the LabWAS for SAMM50 GReX revealed no associations that met the Bonferroni corrected p-value, though nominal signals were seen for hepatic biomarkers like urobilinogen (p=0.03, Figure 2B, Supplemental File 2). Analyses in individuals of African genetic ancestry (n=15,175) showed nominal liver associations but did not surpass multiple-testing correction, likely due to reduced statistical power (Supplemental Files 3–4).
Validation of SAMM50 genetic associations with metabolic conditions and liver disease in additional biobank populations.
BioBank Japan
To validate the BioVU analyses linking SAMM50 GReX to liver disease susceptibility, we next interrogated publicly available summary GWAS data for BioBank Japan. We queried SAMM50 using the BioBank Japan PheWeb platform (https://pheweb.jp/), and found that the SAMM50 genetic locus was significantly associated with liver enzymes (aspartate aminotransferase (AST), and alanine aminotransferase (ALT)), hepatic cancer, liver cirrhosis, type 2 diabetes mellitus (in a multi-ancestry and European metanalysis), and total cholesterol (p < 5.0 × 10^−8^), highlighting SAMM50’s role in metabolic dysfunction and disease in the specific context of hepatic function (Table 2). Furthermore, SAMM50 genetic variation was also significantly associated with abnormalities of platelets and red blood cell indices (p < 5.0 × 10^−8^).
UK Biobank
To validate the BioVU and BioBank Japan analyses linking SAMM50 GReX to liver disease susceptibility, we next interrogated the UK Biobank using publicly available summary GWAS data. We queried SAMM50 using the UK Biobank PheWeb platform^22^(https://pheweb.org/UKB-SAIGE/) and found supporting evidence that the SAMM50 genetic locus was significantly associated with liver phenotypes in this large, independent cohort (Table 2). Specifically, SAMM50 genetic variation is significantly associated with chronic liver disease and cirrhosis, other chronic nonalcoholic liver disease, alcoholic liver damage, portal hypertension, liver abscess and sequelae of chronic liver disease, and cholelithiasis and cholecystitis (p < 5.0×10^−8^).
We then examined the GWAS data from the UK Biobank for further comparison between SAMM50 genetic variation and Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD) (previously referred to as NAFLD, an outdated term), other liver disorders and systemic metabolic traits, Supplementary files 5. Several SAMM50 variants showed strong association with MASLD and related liver pathologies, including rs2143571-A which was associated with nonalcoholic steatohepatitis-derived hepatocellular carcinoma (Study accession GCST005309; OR = 2.7 (95% CI 1.82–4.0), RAF = 0.43, p = 9 × 10^−7^) and MASLD (Study accession GCST005190; OR = 1.437 (95% CI 1.292–1.597) RAF = 0.395, p = 2 × 10^−11^). Other SAMM50 variants, rs2073080-T (GCST001576; OR = 1.47 [95% CI 1.26–1.71], RAF = 0.458, p = 8 × 10^−7^) and rs5764430 (GCST90091033; an effect reported as a 0.1360267-unit decrease, p = 7 × 10^−12^) were associated with MASLD, Supplementary files 5.
We further found that SAMM50 variants were consistently associated with minimum, median, and maximum hepatic fat levels, as measured by imaging-derived hepatic fat traits. The effect sizes for these associations ranged from approximately +1.0 to +1.5 units across multiple entries. For instance, maximum and median hepatic fat were positively associated with a β of +1.451(95% CI 1.11–1.79; p = 6 × 10^−17^, GCST90255386) and β = +1.323 (95% CI 0.96–1.68, p = 5 × 10^−13^, GCST90255381)
We also found strong association of SAMM50 variants with liver enzymes. The variants associated with AST levels included rs7587-T (GCST90019497; β = −0.0394 (95% CI −0.034 to −0.045), p = 2 × 10^−43^), rs12167845-T (GCST90662897; β = −0.0773 (95% CI −0.073 to −0.081), p = 2 × 10^−330^), Supplementary files 5. The same and additional variants were associated with ALT levels including rs7587-T (GCST90019492; β = −0.0402 (95% CI −0.035 to −0.046), p = 2 × 10^−45^), rs2073080-T (GCST90104161; β = −1.53366 IU/L, p = 3 × 10^−10^). rs3761472-G (GCST007440; β = +0.0362 (95% CI 0.02–0.052), p = 7 × 10^−6^). The variant rs7587-T above that was associated with ALT and AST levels was also associated with AST/ALT ratio (GCST90019498; β = +0.0243 [95% CI 0.019–0.030], p = 1 × 10^−17^).
We also found very strong associations between SAMM50 variants and markers of metabolic derangements/disorders such as lipids, uric acid/gout, diabetes and other related conditions. For instance, the variant rs4823182-G shows was associated with type 2 diabetes mellitus (GCST006867; β = +0.0482 [95% CI 0.033–0.063], p = 3 × 10^−10^), hypogonadism (GCST90503324; β = −0.14782159 [95% CI 0.098–0.198], p = 2 × 10^−8^) while the variant rs2235776-C was associated with total cholesterol (GCST003302; β = +0.026 [95% CI 0.016–0.036], p = 3 × 10^−8^). The variant rs7587-T was associated with Triacylglycerol_56:7 (GCST90060551; β = −0.113349 [0.079–0.148], p = 2 × 10^−10^), Triacylglycerol_58:7 (GCST90060564; β = −0.113649 [0.078–0.149], p = 3 × 10^−10^) and Diacylglycerol_38:5 (GCST90060200; β = −0.0767797 [0.057–0.096], p = 1 × 10^−14^) including cholesterol-to-total lipids ratio in large VLDL for the variant rs2294922-G (GCST90092869; β = −0.0337963 [0.024–0.044], p = 1 × 10^−11^). Gout (PheCode 274.1) at rs4823109 was also associated (GCST90651115; β = −0.0815 [95% CI 0.052–0.111], p = 3 × 10^−8^). From the UK biobank, it is evident that SAMM50 is associated with metabolic and hepatic trait enrichment, showing particularly strong signals for liver fat, MASLD/NASH-related phenotypes, liver enzymes, and downstream disease outcomes and severity.
In addition, we note that SAMM50 and PNPLA3 jointly show strong associations with MASLD (p = 1 × 10^−18^), ALT/AST (p = 5 × 10^−24^) and AST (p = 9 × 10^−19^), highlighting a genomic region densely linked to liver fat and enzyme signals. Multiple entries mapping exclusively to SAMM50, such as NAFLD and extreme AST/ALT levels, further support its role in hepatic and metabolic traits.
Nutrient State is associated with Regulation of SAM50 and Respiratory Complex Expression
Next, we aimed to determine how the presence or absence of nutrients relates with the structure and makeup of mitochondria. In this study, we thoroughly examined the expression of mitochondrial respiratory chain protein complexes and the detailed structure of mitochondria in mouse liver following periods of fasting and re-feeding. We also studied cultured cells under serum starvation conditions. We then explored how changes in nutrient levels affect mitochondrial composition.
Immunoblot analysis revealed that re-feeding after a 24-hour fast was significantly associated with increased hepatic protein levels of SAM50 (1.00 vs. 1.93, p=0.044) and Complex V (1.00 vs. 0.61, p=0.022) compared to the fasted state, while Complex I (1.00 vs. 0.41, p=0.0006) and Complex III (1.00 vs. 0.43, p=0.008) were decreased. Protein levels of Tom20 and Complex II (SDHA) were unchanged (Figure 3A–G). In cultured cells, 6-hour serum starvation was associated with mitochondrial enlargement and reduced cristae density by transmission electron microscopy (TEM) (Figure 4A, B, D, E). Immunoblotting of these cells showed that serum starvation was significantly associated with decreased protein levels of Complex II (1.00 vs. 0.43, p=0.018), Complex III (1.00 vs. 0.48, p=0.009), and Complex V (1.00 vs. 0.48, p=0.020), but not Tom20, SAM50, or Complex I (Figure 4C, F–K).
Aging is associated with Remodeling of Hepatic Mitochondrial and Lipid Droplet Architecture
For our aging experiments, we aimed to investigate the relationship between aging and mitochondrial architecture and composition in the liver. We therefore performed 3D volumetric reconstructions of mitochondria from 3-month and 2-year mouse livers using serial block-face scanning electron microscopy (SBF-SEM), followed by quantitative morphometric and molecular analyses.
Using SBF-SEM for 3D reconstruction, we found that aging (2-year vs. 3-month mouse liver) was associated with reduced mitochondrial size, volume, and network complexity and increasing sphericity. Specifically, mitochondrial 3D volume was lower in 2-year (0.66 μm^3^) compared to 3-month old (0.82 μm^3^), p<0.0001, and sphericity was higher in older vs young mice (0.80 vs 0.65, p<0.0001). The mitochondrial complexity index was markedly lower in 2-year vs. 3-month old mice liver (1.39 vs. 2.64, p<0.0001) (Figure 5A–G). This structural change was associated with a ~44% difference with Sam50 mRNA expression lower in 2-year vs 3-month mice (0.60 vs. 1.08, p=0.0019) and lower mitochondrial DNA content (0.75 vs. 0.90, p=0.0008) (Figure 5H, I). SBF-SEM further revealed age-dependent expansion of lipid droplets, with the percentage of lipid coverage within the imaged volume about fourfold higher in 2-year vs. 3-month-old (16.38% vs. 3.72%, p<0.0001) (Figure 6, 7A–E). Aged livers exhibited higher lipid droplet-mitochondria contact site coverage (7.39 vs. 4.10, p=0.0004) and a higher mitochondrial surface area engaged in contacts (7.85% vs. 4.14% vs. p=0.0014) (Figure 7F, G). Biochemically, aging was associated with increased lipid peroxidation, as evidenced by higher plasma TBARS (8.05 vs. 96, p=0.003), higher hepatic triglycerides (0.65 mg/g vs. 0.25, p=0.0002), and a doubling of serum triglycerides (0.32 vs. 0.16 mg/dL, p=0.005). Bile acid levels were unchanged (Figure 7H–M).
High-Fat Diet is associated with Metabolic Dysfunction and Mitochondrial Remodeling Partially Reversible Upon Dietary Reversal
Chronic high-fat diet (HFD) feeding in mice was associated with progressive weight gain and suppressed Sam50 mRNA expression. This suppression was partially restored upon dietary reversal (from HFD to LFD), while the abundance of SAM50 protein remained unchanged across groups (Figure 8A–F). Untargeted metabolomics revealed global hepatic metabolic reprogramming in response to HFD, as evidenced by clear separation in principal component analysis and distinct metabolite signatures (Figure 8G–I).
Discussion
Our study provides evidence for the role of mitochondrial SAM50 as a critical nutrient-related and age-sensitive regulator of mitochondrial function and structure with direct clinical implications that link it to human hepatic and metabolic disorders.
Human Genetic Associations Positioning SAMM50 within Hepatic Disease Pathogenesis
We compared the association between SAMM50 genetic variation and liver-related metabolic and non-metabolic disorders across three large biobanks and found a strong relationship exists. Data from BioVU is rich with clinical and genetic data and our findings show that genetically determined lower SAM50 expression is a significant risk factor for a spectrum of liver diseases, including both metabolic and alcohol-associated etiologies such as chronic nonalcoholic liver disease, cirrhosis, portal hypertension, esophageal bleeding related to variceal disease, and liver transplant status. Our findings are clinically meaningful representing disease phenotypes, severity and progression rather than isolated biochemical abnormalities. Our findings from BioVU aligns with and significantly extends recent reviews implicating mitochondrial outer membrane integrity and protein import in the pathogenesis of metabolic liver disease^9,32^. The specificity of the PheWAS signal for hepatic phenotypes underscores the liver’s particular vulnerability to perturbations in mitochondrial outer membrane assembly. The dissociation between strong disease associations and weaker LabWAS signals suggests that SAM50 deficiency may predispose to structural pathology and long-term disease progression rather than acutely altering circulating biomarkers, aligning with the observed ultrastructural defects in our mechanistic studies and supports a model where chronic, subcellular architectural decay culminates in clinically overt organ dysfunction^33,34^.
The data from BioBank Japan shows the strong association between SAMM50 genetic variation and liver enzymes (ALT and AST) used as biomarkers for liver injury, with ALT being a more liver-specific biomarker. Notably, the strong association between SAMM50 and liver enzymes was very strong, with p-values reaching orders of magnitude beyond conventional GWAS significance thresholds in cohorts exceeding 150,000 individuals. This data suggests that mitochondria SAMM50 plays a critical role in liver-related injury. Additional findings from the BioBank Japan linking SAMM50 with increased odds for type 2 diabetes mellitus, liver cirrhosis, and liver cancer may suggest that the genetic variants of SAMM50 extends beyond subclinical derangements to overt clinical disease. Although beyond the scope of this study, the associations we observed between SAMM50 and several traits including gout, platelet count, and red blood cell indices suggests potential pleiotropic effects of SAMM50 on mitochondrial-dependent metabolic and hematological pathways. Our genetic results are derived from a gene-centered analytical approach, offering validation of SAMM50 as a liver-relevant mitochondrial locus, that is independent of lead variant selection.
The UK Biobank data provides complementary information that increases rigor and strengthens generalizability of our BioVU PheWAS and LabWAS findings. From the UK Biobank data that assesses more than 50 associations, we found that multiple independent variants within the SAMM50 locus show robust and reproducible associations with MASLD, nonalcoholic steatohepatitis and nonalcoholic steatohepatitis derived hepatocellular carcinoma with odds ratios ranging from approximately 1.4 to greater than 2.7. Findings related to the strong association between SAMM50 and ALT and AST are similar between the Biobank Japan and UK Biobank highlighting consistency across multiple populations and supporting the role for SAMM50 in liver-related injury. Using UK Biobank data, we also identified associations between SAMM50 genetic variation with quantitative imaging derived measures of hepatic fat including minimum, median and maximum liver fat content thereby, establishing a direct link between this mitochondrial locus and hepatic steatosis. The UK Biobank studies also reveal extensive SAMM50 associations with lipid metabolism including total cholesterol, triglyceride and diacylglycerol species, type 2 diabetes, gout and related metabolic phenotypes. Although several liver associated signals occur in a genomic region shared with PNPLA3, a well-established fatty liver disease gene, multiple associations map directly to SAMM50 alone supporting an independent contribution of this mitochondrial gene to hepatic and metabolic pathology.
Taken together, the three datasets establish coherent and highly reproducible disease axis linking SAMM50 in humans. SAMM50 encodes a core component of the mitochondrial outer membrane sorting and assembly machinery and is essential for proper mitochondrial membrane organization and cristae integrity. Thus, the consistent associations we have shown from the three large databases provide a strong human genetics foundation for the experimental findings presented in this study which interrogate mitochondrial structure lipid handling and stress responses.
Nutrient effects on Mitochondrial Remodeling
We aimed to elucidate the association between nutrient availability or lack on mitochondrial composition and architecture. In this experiment we critically assessed mitochondrial respiratory chain protein complex expression and ultrastructure in mouse liver that was subjected to fasting and re-feeding, as well as in cultured cells that were exposed to serum starvation. Across both in vivo and in vitro systems, we found an association between nutrient deprivation and coordinated but selective changes in mitochondrial protein abundance and morphology, rather than uniform alterations in mitochondrial content. These feeding/refeeding and fasting experiments reveal a layered regulatory mechanism related to SAM50. Fasting/re-feeding and serum starvation dynamically and selectively remodel respiratory protein complex abundance, with SAM50 protein increasing upon nutrient restoration in vivo. These findings therefore suggest that SAM50 expression may potentially be part of a coordinated anabolic response to nutrients, potentially driving mitochondria bioenergetics^7,35^. The contrasting stability of the SAM50 protein during cellular serum starvation, despite the loss of downstream respiratory protein complexes, suggests a potential buffering mechanism or prioritized maintenance of the import machinery during acute stress, which may possibly facilitate rapid recovery^36^. The underlying role is beyond this study. However, a study by Yin et al demonstrated that in nutrient and oxygen deprivation during ischemic stroke and reperfusion injury in rats, Sam50 played a protective role against injury affecting mitochondria and neurons^37^. Although their study was not directly dealing with starvation, it highlights similar deficiency-driven effects. The selective remodeling of complexes I, III, and V underscores that nutrient states do not uniformly regulate all ETC components, meaning that there could be novel regulatory checkpoints involved. This differential regulation has functional consequences, as the shape and density of cristae, governed by complexes like the MICOS/SAM axis and ATP synthase, directly determine the organelle’s capacity for energy conversion versus metabolite biosynthesis^3^. Our data suggest SAM50 is a key component in translating nutrient signals into these functional structural adaptations.
Aging is associated with Mitochondrial Altered state
To determine the relationship between aging and mitochondrial architecture and composition in the liver, we performed 3D volumetric reconstructions of mitochondria from 3-month and 2-year mouse livers using serial block-face scanning electron microscopy (SBF-SEM), followed by quantitative morphometric and molecular analyses. Our studies demonstrates that aged mice were associated with lower mitochondrial size, perimeter, and volume and higher/increased sphericity. We also found that aging is associated with reduced mitochondrial network complexity, decreased SAM50 expression and mtDNA content, expansion of lipid droplet size and abundance, increased lipid droplet–mitochondria contact site coverage indicating that aging enhances physical coupling between mitochondria and lipid droplets in the liver. In addition, we also found that aging is associated with increased lipid peroxidation in the liver and plasma and alters systemic and hepatic lipid metabolism. Taken together, these results reveal that aging drives coordinated remodeling of mitochondrial and lipid droplet architecture in the liver.
Our findings that aging is associated with mitochondrial dysfunction, mainly characterized by reduced mitochondria size, volume, complexity, and mtDNA content, coupled with a decline in Sam50 expression, is consistent with its identification as a feature of age-related mitochondrial dysfunction^38–41^. The concurrent expansion of lipid droplets and increase in lipid droplet-mitochondria contact sites in aged liver as found in our study presents a compelling model for age-related metabolic decline^38–41^. Enhanced physical coupling, as extensively reviewed in the context of metabolic health, may reflect an adaptive attempt to facilitate fatty acid flux for β-oxidation in the face of declining mitochondrial oxidative capacity^42,43^. However, this increased interface in a milieu of rising oxidative stress, evidenced by elevated lipid peroxidation, may also create a vicious cycle. Mitochondria lipid-droplet in other systems exhibit unique bioenergetics, and our finding suggests that aging may alter the nature of these specialized contacts, potentially shifting them from a source of metabolic support to a source of lipotoxic stress^43^. This may potentially create a vicious cycle of oxidative damage and further mitochondrial dysfunction. More studies are required to elucidate this in much detail
High-Fat Diet and Metabolic Memory: Implications for Intervention
The HFD model encapsulates a pathological acceleration of nutrient-stress responses. HFD was associated with a state of mitochondrial stress evidenced by enlarged, rounded mitochondria with disorganized cristae alongside systemic metabolic syndrome. The persistent associated suppression of Sam50 transcription by HFD, and its only partial recovery after dietary reversal, indicates that chronic metabolic stress can impose a lasting “memory” on mitochondrial regulatory pathways, a phenomenon increasingly recognized in metabolic diseases like MASLD^33,44^. The disconnect between recovered transcription and persistent protein levels and ultrastructure suggests that restoring normal mitochondrial architecture lags behind transcriptional normalization. This has critical implications for the timeline and expectations of metabolic recovery after dietary or therapeutic intervention, implying that strategies to actively promote mitochondrial repair may be necessary to reverse this dysfunction.
SAM50 acts as a key regulator of mitochondrial outer membrane structure and protein import, and is associated with modulation of inner membrane formation, respiratory complex assembly, and oxidative metabolism. Reduced SAM50 expresison, due to genetics, aging, or nutrient overload, is associated with damaged mitochondrial integrity, decreased respiration efficiency and increased oxidative stress, which can impair cells and organs, especially in the liver.
Clinical Implications
The findings from this integrated study carry significant translational implications, bridging fundamental mitochondrial biology to human health. The strong genetic association between SAMM50 expression and liver disease risk, coupled with its dynamic regulation by diet and aging, positions SAM50 as a potential biomarker for metabolic liver disease susceptibility and progression. Quantifiable alterations, such as the mitochondrial simplification and increased lipid droplet-mitochondria contacts identified through advanced microscopy, could evolve into novel histopathological or imaging-based biomarkers for staging steatotic liver disease. Therapeutically, the SAM/MICOS axis emerges as a compelling novel target. Strategies designed to bolster SAM50 function or expression could potentially enhance mitochondrial protein import, cristae integrity, and oxidative metabolism, thereby addressing foundational defects in conditions like MASLD. However, the observed “metabolic memory” effect, where HFD-induced suppression of Sam50 transcription and mitochondrial remodeling were only partially reversed, offers a crucial caution. This suggests that therapeutic benefits may require sustained intervention to overcome entrenched mitochondrial dysfunction and that simply removing the metabolic insult through diet alone might be insufficient for full recovery. Consequently, our finding underscores the paramount importance of early dietary and lifestyle intervention, particularly in aging populations where SAM50 expression is naturally in decline, to prevent the initial entrenchment of mitochondrial pathology.
Strengths of the Study
This study is distinguished by several key methodological strengths that reinforce the validity and impact of its conclusions. Its foremost strength is the translational integration of genetic discovery and mechanism, beginning with data from 3 population-scale genetic studies including a large clinical biobank to establish human disease relevance and then employing controlled in vivo and in vitro models to dissect and assess the underlying cellular pathophysiology. This approach creates a powerful and credible pipeline from population-level genetics to molecular mechanism. Furthermore, our study employs state-of-the-art volumetric imaging, specifically serial block-face scanning electron microscopy, moving beyond the descriptive or conventional 2D snapshots or view and provide three-dimensional quantitative analysis of mitochondrial and lipid droplet morphology which is far more useful. This technique was crucial for rigorously documenting the age-related loss of network complexity and the increase in organelle contact sites. The complementary use of untargeted metabolomics in our study alongside molecular and biochemical assays constitutes a multi-omics framework that connects structural remodeling to global shifts in hepatic metabolic function. Finally, the experimental design thoughtfully models the disease trajectory by examining a spectrum of metabolic states, from acute fasting and natural aging to chronic dietary challenge and attempted reversal, which provides nuanced insights into both the pathogenesis of and potential recovery from metabolic dysfunction.
Limitations
Our study has several limitations. First, the human genetic associations, while robust in the populations of majority European and Asian genetic ancestry (BioVU, UK Biobank, and BioBank Japan), were underpowered in our BioVU studies with individuals of African genetic ancestry, highlighting the need for more diverse biobanks. Second, the in vivo dietary and aging studies are correlative in that while they show strong associations between SAM50 expression, mitochondrial structure, and metabolism, direct causal manipulations of SAM50 in vivo in these specific contexts (aging, HFD) are needed to definitively establish mechanism. Third, the molecular pathways linking nutrient sensing such as mTOR, and AMPK to the transcriptional and post-transcriptional regulation of Sam50 have not been elucidated. Fourth, the functional consequences of the observed increase in lipid droplet-mitochondria contacts in aging require further investigation to determine whether they are adaptive or maladaptive. Finally, the timeframe for the dietary reversal experiment may have been insufficient to observe full normalization of protein levels and ultrastructure; however longer recovery periods should be examined.
In conclusion, our study demonstrates that SAM50 is associated with nutrient state, aging, and diet resulting in changes in mitochondrial structure and systemic metabolism. Firstly, we have demonstrated that SAMM50 genetic variation influences human liver disease risk. Secondly, that SAM50 expression and mitochondrial architecture are dynamically regulated by nutrient availability. Thirdly, that aging is characterized by coordinated loss of SAM50 expression, mitochondrial structural abnormalities, and altered lipid droplet morphology and fourthly, that high-fat diet suppresses Sam50 expression and drives pathological mitochondrial remodeling and metabolic dysfunction, which is only partially reversible upon dietary correction. Our findings position SAM50 as a key regulator and potential target for understanding and potentially intervening in age-related and diet-induced metabolic diseases.
Supplementary Material
Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Chen W, Zhao H, Li Y. Mitochondrial dynamics in health and disease: mechanisms and potential targets. Sig Transduct Target Ther 2023; 8: 333.
- 2Adebayo M, Singh S, Singh AP, Dasgupta S. Mitochondrial Fusion and Fission: The fine-tune balance for cellular homeostasis. FASEB J 2021; 35: e 21620.34048084 10.1096/fj.202100067 RPMC 8415099 · doi ↗ · pubmed ↗
- 3Glancy B, Kim Y, Katti P, Willingham TB. The Functional Impact of Mitochondrial Structure Across Subcellular Scales. Front Physiol 2020; 11. doi:10.3389/fphys.2020.541040. · doi ↗
- 4Anand R, Reichert AS, Kondadi AK. Emerging Roles of the MICOS Complex in Cristae Dynamics and Biogenesis. Biology (Basel) 2021; 10: 600.34209580 10.3390/biology 10070600 PMC 8301002 · doi ↗ · pubmed ↗
- 5Ravi R, Routray D, Mahalakshmi R. Mitochondrial Sorting and Assembly Machinery: Chaperoning a Moonlighting Role? Biochemistry 2025; 64: 312–328.39754567 10.1021/acs.biochem.4c 00727 PMC 7617723 · doi ↗ · pubmed ↗
- 6Höhr AIC, Straub SP, Warscheid B, Becker T, Wiedemann N. Assembly of β-barrel proteins in the mitochondrial outer membrane. Biochimica et Biophysica Acta (BBA) - Molecular Cell Research 2015; 1853: 74–88.25305573 10.1016/j.bbamcr.2014.10.006 · doi ↗ · pubmed ↗
- 7Ott C, Ross K, Straub S, Thiede B, Götz M, Goosmann C Sam 50 Functions in Mitochondrial Intermembrane Space Bridging and Biogenesis of Respiratory Complexes. Mol Cell Biol 2012; 32: 1173–1188.22252321 10.1128/MCB.06388-11PMC 3295012 · doi ↗ · pubmed ↗
- 8Zheng Y, Wang S, Wu J, Wang Y. Mitochondrial metabolic dysfunction and non-alcoholic fatty liver disease: new insights from pathogenic mechanisms to clinically targeted therapy. J Transl Med 2023; 21: 510.37507803 10.1186/s 12967-023-04367-1PMC 10375703 · doi ↗ · pubmed ↗