Introduction
Metabolic dysfunction-associated steatotic liver disease (MASLD) represents a chronic liver condition closely linked to systemic metabolic abnormalities. Its diagnostic criteria have evolved from the exclusion-based “nonalcoholic fatty liver disease” (NAFLD)to a positive framework requiring hepatic steatosis (confirmed by imaging or biopsy) and at least one cardiometabolic risk factor, better capturing the multisystem pathophysiology of the disease. Among those with type-2 diabetes, prevalence rises to 58.84% in East Asian and 72.65% in Western populations. Sedentary lifestyles and dietary shifts away from traditional patterns towards high-fat/sugar intake contribute to this by promoting gut dysbiosis. Although often asymptomatic and incidentally detected through abnormal liver tests or imaging, MASLD carries significant risks of progression to metabolic dysfunction-associated steatohepatitis (MASH) and fibrosis, alongside increased hepatocellular carcinoma incidence in early-stage disease and gallstone formation from supersaturated bile. Hepatolithiasis risk is notably elevated in patients with biliary tract malformations. Cardiovascular disease (CVD) is the main cause of death in MASLD. Hepatic fat, insulin resistance, endothelial dysfunction, and systemic inflammation drive this risk, leading to accelerated atherosclerosis, diastolic dysfunction, and arrhythmias. [] 1 [,] 2 3 [,] 1 4 [] 5 [] 6 [] 7 [,] 8 9
Current MASLD management focuses on lifestyle changes: Mediterranean diet and ≥150 min/week moderate exercise. Weight loss of ≥3–5% improves steatosis and ≥10% reverses MASH/fibrosis. Glucagon-like peptide-1 receptor agonists and sodium-glucose cotransporter-2 inhibitors offer triple benefits: glycemic control, weight loss, and cardiovascular protection. Emerging agents, including peroxisome proliferator-activated receptor (PPAR) agonists and thyroid hormone receptor-β agonists, further show histological improvements in MASH. However, clinical challenges persist: specific, noninvasive methods are still lacking to detect mild steatosis and to accurately identify at-risk groups using tools like the Fibrosis-4 index or abdominal ultrasound. Conventional cardiovascular risk models tend to underestimate MASLD as an independent risk factor. Additionally, gaps persist in clinician awareness and interdisciplinary care coordination. Furthermore, the mechanistic heterogeneity and variable therapeutic responses observed in lean MASLD patients warrant further investigation. [] 4 [,] 10 11 [,] 2 12 [] 13 [] 9 [] 14 [] 15
The triglyceride-glucose index (TyG) serves as a simple yet robust marker of insulin resistance. Elevated TyG correlates with symptomatic coronary artery disease severity and predicts major adverse cardiovascular events in spatial transcriptomics (ST) elevated myocardial infarction patients. It also associates with arterial stiffness, renal microvascular damage, and incident diabetes, highlighting its utility in multisystem metabolic risk assessment. MASLD and TyG are driven by insulin resistance and dyslipidemia. Hepatic fat and systemic inflammation in MASLD promote endothelial dysfunction and atherosclerosis, raising CVD risk. TyG, reflecting hepatic insulin resistance and lipotoxicity, may identify MASLD patients at heightened risk of steatosis progression and fibrosis. Its sensitivity might improve risk stratification in populations that are underestimated by traditional models. [,] 16 17 [] 18 [,] 19 20 [,] 8 9 [] 14 [,] 13 15
Mendelian randomization (MR) is a key method for studying disease causes, especially when randomized controlled trials are not possible. It uses single-nucleotide polymorphisms (SNPs) strongly linked to exposures as instrumental variables (IVs) to mimic RCT randomization via the genetic lottery at conception. This approach inherently minimizes confounding biases, as genetic variants remain unaffected by postnatal factors like age or sex. Transcriptome-wide association studies (TWAS) integrate expression quantitative trait loci (eQTL) data with genome-wide association study (GWAS) summary statistics to identify candidate genes and investigate gene–trait associations. Methods such as Multi-marker Analysis of GenoMic Annotation (MAGMA)and Functional Summary-based Imputation (FUSION)are employed. [,] 21 22 [] 23 [] 24 [] 25 [] 26
The primary objective of this study is to investigate the association between the TyG index and MASLD through MR and genomic analyses, utilizing 192 strongly correlated SNPsas proxies for TyG. This approach aims to identify shared genetic pathways and advance integrated strategies for cardiovascular-liver-metabolic health (CLMH). The significance of this work lies in three key aspects. First, the TyG index may serve as a complementary tool for early MASLD screening by identifying individuals with insulin resistance and metabolic abnormalities, thereby improving the sensitivity of noninvasive diagnostic methods. Second, combining TyG with hepatic fat quantification could refine cardiovascular risk stratification in MASLD patients, enabling personalized management. Furthermore, longitudinal monitoring of TyG dynamics may evaluate the efficacy of lifestyle modifications or pharmacological interventions in improving metabolic and cardiovascular outcomes. In accordance with the TITAN Guidelines 2025 (Supplemental Digital Content Materials, available at:) governing the declaration and use of artificial intelligence, we confirm that no AI tools were utilized in the data generation, analysis, or manuscript preparation of this study. [] 27 [] 28 http://links.lww.com/JS9/F272
Methods
We initially collected 192 strongly associated proxy SNPs for the TyG index from a UK Biobank cohort study with genetic risk scores, followed by imputation using the GRCh37 reference genome from the 1000 Genomes Project European population to generate chromosome positions, base pair locations, and effect allele frequencies (EAF). MR was applied to evaluate causal relationships between TyG and MASLD using these SNPs. Transcriptomic analyses integrated four complementary approaches: (1) MAGMA for gene-set enrichment analysis; (2) Joint-Tissue Imputation Enhanced PrediXcan (JTI-PrediXcan) to predict tissue-specific gene expression; (3) Summary-based Multi-Tissue Imputation Xcan (SMulTiXcan) to aggregate cross-tissue expression signals; and (4) Fine-mapping of Causal Gene Sets (FOCUS) to prioritize causal genes via Bayesian fine-mapping. Consensus genes were defined by concordance across two or more methods. FUSION was independently applied to validate shared gene–MASLD associations through expression-trait colocalization. Candidate genes were further prioritized using the Polygenic Priority Score (PoPS), which integrates functional genomic annotations and pleiotropic evidence to rank genes with robust associations. Finally, single-cell spatial transcriptomics (sc-ST) was integrated with MASLD GWAS data, leveraging embryonic mouse E16.5 ST datasets spanning 25 organs to systematically map MASLD-associated cellular architectures and organ-specific distributions of shared genes at single-cell resolution. [] 27 [] 25 [,] 29 30 [,] 29 31 [] 32 [] 26 [] 33 [] 34
Flow chart of our study. GWAS, Genome-Wide Association Study; TyG, triglyceride-glucose index; MASLD, metabolic dysfunction-associated steatotic liver disease; SNPs, single nucleotide polymorphisms; GRS, genetic risk scores; TWAS, transcriptome-wide association study; GTEx, Genotype-Tissue Expression Project; gsMap, genetically informed spatial mapping; MR, Mendelian randomization; IVW, inverse variance weighted; GSMR, Generalized Summary-data Mendelian Randomization; LD, linkage disequilibrium; MAF, Minor Allele Frequency; MAGMA, Multi-marker Analysis of GenoMic Annotation; JTI-PrediXcan, Joint-Tissue Imputation PrediXcan; SMulTiXcan, Sparse Multi-Tissue Imputation Xcan; FOCUS, Fine-mapping of Causal Gene Sets; FUSION, Functional Summary-based Imputation; PoPS, Polygenic Priority Score; STRING, Search Tool for Recurring Instances of Neighboring Genes; gnomAD, Genome Aggregation Database; pLI, Probability of Loss-of-function Intolerance; ENCODE, Encyclopedia of DNA Elements; GSS, Gene-Specific Score.
Sources of TyG and MASLD data
In the Methodology section about MASLD data analysis, we utilized the 2025 updated dataset derived from a comprehensive GWAS study employing UK Biobank resources. This study cohort comprised 501 289 participants with documented self-reported sex and birth dates, enabling rigorous evaluation of verified MASLD case-control status. GWAS summary statistics for the TyG index were acquired from the UK Biobank, a population-based cohort comprising 273 368 European-ancestry participants aged 40–69 years, excluding individuals with pre-existing diabetes mellitus (DM) or lipid metabolism disorders. IV selection followed stringent criteria: SNPs achieving genome-wide significance (< 5 × 10) were identified via linear regression models adjusted for age, sex, and the top five genetic principal components to mitigate population stratification. Subsequent clumping (clump< 0.01, 10 000kb window) and exclusion of SNPs associated with glucose (GLU) or triglyceride (TG) levels yielded 192 independent TyG-associated SNPs for downstream analyses. The above SNPs were interpolated using the human reference genome GRCH37 reference file from the European 1000 Genomes Project to generate chromosome and base pair location and EAF. The full list of 192 SNPs is provided in Supplemental Digital Content Table S2, available at:. During GRS construction for TyG indices, individuals with diabetes mellitus and dysregulated lipoprotein metabolism were excluded. SNPs were stringently filtered, removing those associated with TG/GLU levels or correlated with non-lipid/non-glycemic traits: systolic blood pressure, diastolic blood pressure, and body mass index. These methodological safeguards effectively mitigated potential confounding effects and substantially addressed sample overlap concerns inherent in UK Biobank-derived datasets. [] 36 [] 27 −8 2 [] 27 P R http://links.lww.com/JS9/F463
Statistical analysis
MR analysis
A valid MR study must satisfy three core assumptions: (1) a strong and robust association between IVs and the exposure; (2) independence of IVs from confounders affecting the exposure-outcome relationship; and (3) genetic variants influencing the outcome exclusively through the exposure, without alternative pathways. We employed six MR approaches to validate causal relationships between TyG and MASLD: inverse variance weighted (IVW), MR-Egger, weighted median, weighted mode, simple mode, and generalized summary-data MR (GSMR). The IVW method is considered the most accurate estimator under valid instrumental assumptions and serves as our primary analytical framework. GSMR, implemented via the R package developed by Zhu, enhanced causal inference by integrating pleiotropy-robust pruning and heterogeneity adjustment. [] 37 [,] 38 39 [,] 40 41 et al
Sensitivity analysis
Sensitivity analyses were conducted to assess violations of horizontal pleiotropy and heterogeneity in MR estimates. Horizontal pleiotropy was evaluated via MR-Egger regression, with an intercept test< 0.05 indicating significant pleiotropic bias. Non-significant pleiotropy (≥ 0.05) supported the validity of IVs. Heterogeneity was quantified using Cochran’s-test, where< 0.05 denoted substantial between-SNP heterogeneity. For analyses with significant heterogeneity (-test< 0.05), random-effects IVW models were applied; otherwise, fixed-effects IVW models were employed. P P Q P Q P [] 42 [] 43 [] 44
Comparative evaluation of four TWAS frameworks for TyG and MASLD
We integrated MAGMA due to its robust framework for gene and gene-set association testing. It aggregates SNP-level signals (< 5 × 10) into gene-level associations using principal component regression, which explicitly models linkage disequilibrium (LD) and mitigates confounding by gene size. This approach detects multi-marker effects often missed by SNP-wise methods. MAGMA’s two-tiered architecture – separating gene analysis from gene-set testing – ensures flexibility and statistical efficiency, enabling reliable identification of biological pathways. We developed JTI-PrediXcan to integrate cross-tissue regulatory similarity (expression and epigenomic profiles) through elastic net optimization, training multi-tissue prediction models on GTEx v8 eQTL weights across 49 tissues. This framework enhances association power by borrowing information from biologically related tissues, particularly for underpowered tissues. SMulTiXcan enhances cross-tissue gene detection by meta-analyzing tissue-specific Summary-PrediXcan (S-PrediXcan) results. It leverages Multi-trait Adaptive Shrinkage (MASHR) effect models from GTEx v8 and corrects inter-tissue correlation via a SNP covariance matrix. This boosts power to detect genes with distributed or synergistic effects across tissues, addresses single-tissue limitations, and reduces false positives. FOCUS employs a Bayesian fine-mapping framework to probabilistically prioritize causal genes within GWAS risk loci. It addresses a key limitation of standard TWAS: LD induces spurious gene–trait associations at noncausal genes near causal variants. By modeling the correlation structure of TWAS signals – driven by LD and eQTL weights – FOCUS computes posterior inclusion probabilities (PIPs) and generates ρ-credible gene sets. These sets represent genes with a defined probability of containing the causal gene(s) explaining the association signal. Consensus genes required agreement across two or more TWAS methods. When different methods – each with distinct underlying assumptions and approaches (MAGMA’s gene-level signal aggregation, JTI-PrediXcan/SMulTiXcan’s tissue-specific modeling, and FOCUS’s causal fine-mapping) – converge on the same gene–trait association, it strongly suggests that the finding is robust and less likely to be a technical artifact or false positive specific to any single method’s limitations. This concordance effectively mitigates the individual weaknesses of each approach and provides independent validation, collectively ensuring the identification of biologically plausible gene targets with higher confidence.-values were adjusted using the false discovery rate (FDR) method to account for potential correlations among phenotypes, as the Bonferroni correction was deemed overly conservative. The final set combines TyG- and MASLD-associated genes meeting this reproducibility threshold, highlighting loci linked to both traits. Detailed methods are in Supplemental Digital Content Table S1, available at:. P P −8 [] 25 [] 25 [,] 29 30 [] 45 [] 46 [,,] 29 45 46 [] 32 [] 32 [,,] 32 47 48 [] 49 http://links.lww.com/JS9/F463
Independent validation of comorbidity genes using FUSION TWAS framework for MASLD pathogenesis
To confirm the robustness of TyG-MASLD comorbidity genes identified through multi-method consensus, we performed independent validation using the FUSION TWAS platform on the MASLD GWAS dataset. The FUSION framework (Supplemental Digital Content Table S1, available at:) was employed to systematically integrate GWAS summary statistics with tissue-specific eQTL data, enabling TWAS to identify genes mechanistically linked to MASLD. This platform leverages multiple predictive models to estimate SNP-based weights for gene expression, thereby capturing polygenic effects while addressing tissue heterogeneity through cross-tissue meta-analyses. For discovery-phase analyses, genome-wide significant SNPs (< 5 × 10) were prioritized, followed by LD clumping (< 0.001) to ensure independence of IVs. Stringent thresholds were applied to control false discoveries, including FDR for multi-test correction. FUSION enables orthogonal validation by using summary data and reference panels, improving biological interpretability while reducing confounding. http://links.lww.com/JS9/F463 [] 26 [] 26 −8 2 [] 26 P r
Validation of TyG-MASLD comorbidity genes using PoPS framework
We validated TyG-MASLD comorbidity genes using PoPS, which integrates GWAS data with multi-omics to prioritize causal genes. PoPS (Supplemental Digital Content Table S1, available at:) leverages polygenic enrichments across 57 543 gene features – including single-cell RNA-seq expression profiles, pathway annotations, and protein–protein interaction networks – to assign priority scores reflecting shared functional characteristics of causal genes. Key strengths: tissue-specific modeling, reduced false positives via L2 regularization, and genome-wide prioritization. We applied PoPS to TyG/MASLD data using genome-wide significant SNPs (< 5 × 10) and ancestry-matched LD panels. Its cross-trait architecture integration – validated at 74% precision – identified shared pathogenic mechanisms. Intersecting PoPS-prioritized genes with consensus loci validated targets with convergent transcriptomic evidence. This resolved the pleiotropy/tissue heterogeneity limitations of single-method approaches. http://links.lww.com/JS9/F463 [] 33 [] 33 −8 [] 33 P
Single-cell spatially resolved transcriptomic characterization of TyG-MASLD shared genes in MASLD
To systematically investigate the co-pathogenic genes and tissue-specific spatial distribution patterns linking the TyG index and MASLD, this study employed the genetically informed spatial mapping (gsMap) method, published in Nature in 2025. This approach integrates sc-ST data with GWAS statistics for MASLD. The gsMap (Supplemental Digital Content Table S1, available at:) framework leverages a graph neural network (GNN) to harmonize gene expression profiles, spatial coordinates, and GWAS-derived trait associations. It evaluates heritability enrichment using stratified linkage LD score regression (S-LDSC) and quantifies disease-region associations via the Cauchy combination test. By overcoming the spatial resolution limitations of conventional single-cell RNA sequencing, gsMap enables precise mapping of spatially resolved cell populations associated with complex traits. Using mouse embryos (E16.5, 25 organs) and human GWAS data, we generated single-cell spatial maps of TyG-MASLD co-pathogenic genes. Cross-species validation provided spatial multiomics evidence for TyG-MASLD comorbidity mechanisms. [] 34 [] 34 [] 34 http://links.lww.com/JS9/F463
Results
MR analysis and sensitivity analysis results
MR analysis revealed a significant causal association between the TyG and MASLD (Supplemental Digital Content Table S3, available at:). Using the IVW method, we observed a positive causal effect (OR = 1.58, 95% CI: 1.04–2.38,= 0.030), supported by consistent findings from GSMR analysis (OR = 1.43, 95% CI: 1.27–1.61,= 5.20 × 10). Sensitivity analyses identified significant heterogeneity across IVs (Cochran’s= 1.88 × 10), prompting the use of a multiplicative random-effects IVW model, which retained significance (OR = 1.58, 95% CI: 1.04–2.38,= 0.030). Horizontal pleiotropy was not detected (MR-Egger intercept= 0.102), confirming the robustness of the causal estimates. These results underscore TyG as a genetically driven risk factor for MASLD, independent of confounding pathways. http://links.lww.com/JS9/F463 P P Q P P P −9 −42
Comparative evaluation of four TWAS frameworks for TyG and MASLD
MAGMA-based tissue-specific enrichment and pathway enrichment analyses for TyG and MASLD. This composite figure presents MAGMA results comparing TyG (top row) and MASLD (bottom row). Left panels show tissue-specific enrichment: TyG was significantly enriched in adipose, vascular tissues, and liver, while MASLD showed enrichment in gastrointestinal and kidney tissues. Right panels show the top enriched biological pathways: TyG pathways centered on lipid metabolism and stress responses, whereas MASLD pathways are involved in developmental morphogenesis and immune signaling. These distinct enrichment patterns highlight the differing tissue involvements and underlying biological mechanisms between the TyG index and MASLD. TyG, triglyceride-glucose index; MASLD, metabolic dysfunction-associated steatotic liver disease.
Manhattan Plots of MAGMA-Based Positive Gene Selection for TyG (Left) and MASLD (Right). Genes surpassing the FDR-adjusted significance threshold (< 0.05) are highlighted. Key metabolic regulators TM6SF2 and GCKR exhibit genome-wide significance in both traits, linking insulin resistance (TyG) to hepatic lipid dysregulation (MASLD). Notably, PNPLA3 is uniquely prioritized in MASLD, underscoring its established role in steatosis progression. Vertical dashed lines indicate chromosome boundaries. TyG, triglyceride-glucose index; MASLD, metabolic dysfunction-associated steatotic liver disease. P
Shared candidate genes between TyG and MASLD identified by transcriptomic analysis. The Venn diagram illustrates the 12 comorbid genes consistently identified by at least two independent transcriptomic methodologies. Variation in the number of supporting methods per gene (detailed in Table) reflects differences in tissue-specific expression patterns and underlying genetic effect sizes. Genes validated by more frameworks (eg, TM6SF2, GCKR) typically exhibit stronger associations and broader tissue relevance, highlighting their heightened biological significance in the TyG-MASLD comorbidity pathway. TyG, triglyceride-glucose index; MASLD, metabolic dysfunction-associated steatotic liver disease. 1
| Gene | Methods (TyG) | Methods (MASLD) |
|---|---|---|
| C2orf16/SPATA31H1 | JTI-PrediXcan, SMulTiXcan | MAGMA, JTI-PrediXcan, SMulTiXcan |
| FNDC4 | JTI-PrediXcan, SMulTiXcan | JTI-PrediXcan, SMulTiXcan |
| GCKR | MAGMA, JTI-PrediXcan, SMulTiXcan | MAGMA, FOCUS, JTI-PrediXcan |
| GMIP | JTI-PrediXcan, SMulTiXcan | MAGMA, FOCUS, JTI-PrediXcan, SMulTiXcan |
| HAPLN4 | JTI-PrediXcan, SMulTiXcan | MAGMA, FOCUS, JTI-PrediXcan, SMulTiXcan |
| LPAR2 | JTI-PrediXcan, SMulTiXcan | FOCUS, JTI-PrediXcan, SMulTiXcan |
| MAU2 | JTI-PrediXcan, SMulTiXcan | MAGMA, FOCUS, JTI-PrediXcan |
| MEF2B | JTI-PrediXcan, SMulTiXcan | MAGMA, SMulTiXcan |
| NDUFA13 | JTI-PrediXcan, SMulTiXcan | FOCUS, JTI-PrediXcan |
| NRBP1 | JTI-PrediXcan, SMulTiXcan | FOCUS, JTI-PrediXcan, SMulTiXcan |
| TM6SF2 | MAGMA, SMulTiXcan | MAGMA, FOCUS, JTI-PrediXcan, SMulTiXcan |
| ZNF513 | JTI-PrediXcan, SMulTiXcan | FOCUS, JTI-PrediXcan, SMulTiXcan |
Independent validation of comorbidity genes using FUSION TWAS framework for MASLD pathogenesis
Using the FUSION method for independent validation of TyG-MASLD comorbid genes, nine genes (GMIP, HAPLN4, LPAR2, MAU2, MEF2B, NDUFA13, NRBP1, TM6SF2, and ZNF513) were successfully validated with significantvalues (Supplemental Digital Content Table S16, available at:). Specifically, GMIP showed associations in the Brain Amygdala (= 0.0356) and Minor Salivary Gland (= 0.0440). HAPLN4 exhibited significant signals in the Colon Sigmoid (= 0.0344) and Minor Salivary Gland (= 9.78 × 10). LPAR2 was validated in the Brain Amygdala (= 8.22 × 10). MAU2 demonstrated significance in Adipose Subcutaneous tissue (= 6.97 × 10). MEF2B was associated in the Brain Frontal Cortex BA9 (= 2.60 × 10). NDUFA13 showed a significant signal in the Brain Caudate basal ganglia (= 8.34 × 10). NRBP1 was validated in the Adrenal Gland (= 0.0148). TM6SF2 exhibited significance in the Brain Caudate basal ganglia (= 8.25 × 10), and ZNF513 was associated in the Brain Caudate basal ganglia (= 0.0014). P P P P P P P P P P P P fdr fdr fdr fdr fdr fdr fdr fdr fdr fdr fdr fdr http://links.lww.com/JS9/F463 −9 −7 −5 −8 −7 −7
When evaluating the magnitude of associations (absolute PoPS scores), TyG exhibited markedly stronger polygenic signals compared to MASLD. In TyG, TM6SF2 (7.2491) and GCKR (6.7102) demonstrated exceptionally high absolute scores (>5), followed by FNDC4 (0.7042), GMIP (0.7003), MEF2B (0.7089), and NDUFA13 (0.3693), all exceeding 0.3. In contrast, MASLD-associated scores were uniformly low, with NDUFA13 (0.5028) and MEF2B (0.2001) ranking highest but still below 0.6, followed by LPAR2 (0.2931), C2orf16/SPATA31H1 (0.2745), MAU2 (0.2123), FNDC4 (0.2012), and HAPLN4 (0.1853), whereas TM6SF2 (0.0416) and GCKR (0.0588) showed minimal magnitudes.
Ranking prioritization based on absolute scores further emphasized cross-trait differences. For TyG, the order was: (1) TM6SF2 (7.2491), (2) GCKR (6.7102), (3) MEF2B (0.7089), (4) FNDC4 (0.7042), (5) GMIP (0.7003), (6) NDUFA13 (0.3693), (7) MAU2 (0.4518), (8) ZNF513 (0.3129), (9) LPAR2 (0.2991), (10) C2orf16/SPATA31H1 (0.2345), (11) NRBP1 (0.0609), and (12) HAPLN4 (0.003). In MASLD, the hierarchy was: (1) NDUFA13 (0.5028), (2) LPAR2 (0.2931), (3) C2orf16/SPATA31H1 (0.2745), (4) NRBP1 (0.3126), (5) MAU2 (0.2123), (6) FNDC4 (0.2012), (7) MEF2B (0.2001), (8) HAPLN4 (0.1853), (9) ZNF513 (0.1206), (10) GCKR (0.0588), (11) TM6SF2 (0.0416), and (12) GMIP (0.0366). Key divergences included TM6SF2 and GCKR, which dominated TyG rankings (1st and 2nd) but fell to the lowest positions in MASLD (11th and 10th), and NDUFA13, which rose from 6th in TyG to 1st in MASLD. MEF2B maintained moderate priority (3rd in both), whereas FNDC4 and MAU2 showed comparable mid-tier ranks (4th/6th and 7th/5th, respectively).
| Gene | PoPS score (TyG) | PoPS score (MASLD) |
|---|---|---|
| C2orf16/SPATA31H1 | 0.2345 | −0.2745 |
| FNDC4 | 0.7042 | −0.2012 |
| GCKR | 6.7102 | 0.0588 |
| GMIP | −0.7003 | 0.0366 |
| HAPLN4 | −0.0030 | −0.1853 |
| LPAR2 | −0.2991 | −0.2931 |
| MAU2 | −0.4518 | −0.2123 |
| MEF2B | −0.7089 | 0.2001 |
| NDUFA13 | −0.3693 | −0.5028 |
| NRBP1 | 0.0609 | −0.3126 |
| TM6SF2 | 7.2491 | −0.0416 |
| ZNF513 | 0.3129 | −0.1206 |
Single-cell spatially resolved transcriptomic characterization of TyG-MASLD shared genes in MASLD
Organs such as cartilage primordium (sc-ST= 0.0037; Cauchy= 0.6662) and cavity (sc-ST= 0.0001; Cauchy= 0.9778) further exemplified this contrast, where localized signals lacked organ-wide statistical support. Conversely, epidermis (Cauchy= 0.0995) and submandibular gland (Cauchy= 0.1275), though not prominent in sc-ST mapping, showed moderate but nonsignificant Cauchy-values. These findings emphasize the complementary utility of sc-ST (for spatial hotspots) and Cauchy testing (for systemic associations). P P P P P P P
Cross-species validation of TyG-MASLD comorbidity genes via spatially resolved transcriptomics. The Venn diagram identifies evolutionarily conserved genes (APOA1, APOB, and APOC4) detected by gsMap analysis in embryonic mouse tissues (E16.5), despite the absence of the original 12 human comorbid genes. These apolipoproteins – critical regulators of lipid metabolism – emerged as spatially enriched hubs in liver/vascular microenvironments, highlighting their conserved role in hepatic lipid trafficking and systemic metabolic dysregulation. Their validation underscores species-invariant pathways as high-priority therapeutic targets. GsMap, genetically informed spatial mapping; TyG, triglyceride-glucose index; MASLD, metabolic dysfunction-associated steatotic liver disease.
Spatial expression and GSS of MASLD-associated and TyG comorbid genes in mouse embryos mapped via gsMap (Supplemental Digital Content Materials, available at:). GsMap, genetically informed spatial mapping; TyG, triglyceride-glucose index; MASLD, metabolic dysfunction-associated steatotic liver disease; GSS, Gene-Specific Scores. This figure illustrates the spatial expression profiles and GSS distributions of genes associated with MASLD and comorbid with TyG in embryonic mouse tissues. The analysis, performed using the gsMap method, highlights regions of high genetic activity (spots) where these genes exhibit significant expression. GSS values reflect localized gene expression rankings, emphasizing spatial heterogeneity in metabolic pathways relevant to MASLD and TyG comorbidity. http://links.lww.com/JS9/F420
‘Spatial mapping of MASLD-associated cellular patterns in E16.5 mouse embryonic single-cell spatial transcriptomics data, generated by the gsMap algorithm across 25 organs (Supplemental Digital Content Materials, available at:). GsMap of cells for complex traits. GsMap, genetically informed spatial mapping; MASLD, metabolic dysfunction-associated steatotic liver disease. http://links.lww.com/JS9/F419
| Annotation | Pcauchy | Pmedian |
|---|---|---|
| Adrenal gland | 0.0064 | 0.0219 |
| GI tract | 0.0532 | 0.1472 |
| Connective tissue | 0.0745 | 0.3421 |
| Liver | 0.0797 | 0.1646 |
| Lung | 0.0833 | 0.2163 |
| Epidermis | 0.0995 | 0.23 |
| Submandibular gland | 0.1275 | 0.2783 |
| Adipose tissue | 0.1488 | 0.3808 |
| Kidney | 0.1987 | 0.3033 |
| Mucosal epithelium | 0.3043 | 0.4404 |
| Muscle | 0.3116 | 0.4722 |
| Smooth muscle | 0.3224 | 0.3842 |
| Cartilage | 0.352 | 0.4277 |
| Meninges | 0.3547 | 0.4464 |
| Inner ear | 0.4531 | 0.4956 |
| Choroid plexus | 0.5987 | 0.5566 |
| Bone | 0.6094 | 0.5017 |
| Jaw and tooth | 0.6173 | 0.5499 |
| Cartilage primordium | 0.6662 | 0.5643 |
| Sympathetic nerve | 0.6828 | 0.5739 |
| Dorsal root ganglion | 0.715 | 0.6323 |
| Heart | 0.9642 | 0.6286 |
| Brain | 0.9727 | 0.9237 |
| Cavity | 0.9778 | 0.4556 |
| Spinal cord | 0.99 | 0.9315 |
Discussion
Integrative genomics identified causal links and shared mechanisms between the TyG index and MASLD. MR using 192 SNPs established causality: genetically predicted TyG increased MASLD risk. Multi-method transcriptomics (MAGMA, JTI-PrediXcan, SMulTiXcan, and FOCUS) identified 12 consensus comorbidity genes (e.g., TM6SF2 and GCKR), validated independently (FUSION and PoPS) despite trait directional heterogeneity. Embryonic mouse sc-ST analysis revealed that while these 12 human genes lacked significant expression, three evolutionarily conserved lipid metabolism genes (APOA1, APOB, and APOC4) emerged as cross-species candidates, highlighting conserved metabolic dysregulation pathways.
Our study bridges a critical gap left by prior epidemiological work focused on observational TyG-MASLD associations. Our genetic approach confirms TyG’s multiorgan relevance – supported by clinical evidence of vascular atherosclerosisand gut-liver axis involvement– while revealing key distinctions. Unlike hypertension studies linking TyG to kidney disease, our MR demonstrates that renal signals are secondary to systemic inflammation/renin-angiotensin-aldosterone system (RAAS) activation. Further diverging from cumulative biomarker innovations, our eQTL network identifies APOB-modulated lipid trafficking as an upstream genetic trigger. Our spatial mapping reveals adrenal glucocorticoid-catecholamine dysregulation within pro-fibrotic connective tissue microenvironments – a finding unattainable via circulating biomarkers. Methodological distinctions explain discrepancies: TWAS captures dynamic gene–environment interactions (e.g., GCKR-diet), single-cell resolution reveals occult crosstalk, and genetic instruments mitigate confounders. This transcends correlative paradigms, establishing TyG as a causal driver of MASLD-CVD comorbidity, revealing metabolic vulnerability priming and enabling genotype-guided therapy. TyG index is a clinically actionable, low-cost screening tool for early MASLD detection, especially where conventional models fail. Identified causal genes and apolipoprotein pathways offer theranostic targets; genotyping refines CVD risk stratification and prioritizes patients for targeted therapies. Key challenges include validating tissue-specific therapeutic windows, integrating polygenic risk scores, and developing TyG-genetic screening. Clinicians should implement longitudinal TyG monitoring, and genotype-stratified trials are needed. Integrating TyG-associated genetic variants with established noninvasive fibrosis biomarkers refines MASLD risk stratification by identifying individuals with heightened genetic susceptibility. This integrated approach enhances conventional screening tools, enabling earlier confirmatory imaging and prioritization for targeted therapies, optimizing precision prevention in high-risk cohorts. [] 50 [] 51 [] 52 [] 53 [] 54 [] 55 [] 56
The TyG, a surrogate marker of insulin resistance, contributes to the development and progression of MASLD via multiple mechanisms. Insulin resistance increases hepaticlipogenesis while suppressing fatty acid β-oxidation, leading to abnormal lipid accumulation in hepatocytes. Elevated TyG levels are also associated with systemic inflammation and enhanced oxidative stress, exacerbating hepatocellular injury and fibrosis. Furthermore, TyG elevation directly promotes atherosclerosis via endothelial dysfunction and lipotoxicity, indirectly amplifying cardiovascular risks in MASLD patients. Genetic studies demonstrate that TyG-associated genes influence hepatic lipid deposition by regulating lipid metabolism and gluconeogenic pathways. Clinical evidence indicates that TyG positively correlates with MASLD severity and serves as a biomarker for predicting fibrosis progression. In animal models, TyG elevation promotes hepatic steatosis by activating PPARγ and Sterol Regulatory Element-Binding Protein 1c (SREBP-1c) pathways. Collectively, the causal relationship between TyG and MASLD involves multifactorial interactions encompassing insulin resistance, lipotoxicity, inflammation, and genetic regulation. de novo [] 16 [] 18 [] 8 [] 27 [] 20 [] 19
C2orf16/SPATA31H1’s function remains unknown. GWAS suggests it regulates lipid metabolism via nearby elements, influencing TyG and liver fat. PoPS reveals opposing effects: positive with TyG but negative with MASLD. This paradox likely arises as the gene desert-located C2orf16 acts tissue-specifically, impacting metabolic pathways differently in insulin-sensitive vs. steatotic tissues. FNDC4, a fibronectin domain protein, likely regulates insulin sensitivity via fat browning or inflammation, with levels linked to liver fat. Its positive TyG PoPS score (0.7042) aligns with adipocyte roles, whereas the negative MASLD score (−0.2012) suggests disease downregulation, possibly mitigating inflammatory damage. This shift mirrors known changes in fat cell signaling during NAFLD. GCKR, encoding glucokinase regulatory protein, regulates hepatic GLU metabolism and TG synthesis, and its rs1260326 polymorphism is strongly associated with elevated TyG index and increased MASLD risk. Critically, GCKR has a strong positive PoPS score in TyG (6.7102) and a weak positive score in MASLD (0.0588), indicating a consistent effect promoting metabolic dysregulation. This reflects GCKR’s central role in liver GLU sensing and fat production, driving both hyperglycemia/hypertriglyceridemia (raising TyG) and liver fat buildup in MASLD. GMIP, a Rho GTPase regulator, may disrupt lipid homeostasis by affecting hepatocyte polarity and lipid droplet transport. PoPS reveals a significant direction reversal for GMIP: negative in TyG (−0.7003) but positive in MASLD (0.0366), suggesting opposing functions in different disease stages. In early insulin resistance, reduced GMIP activity may impair cellular structure and fat movement, promoting harmful fat buildup. Conversely, in established MASLD, increased GMIP activity might represent an adaptive response, potentially restoring liver cell function or enabling fat storage, similar to cellular adaptations seen under fat stress. HAPLN4 was identified as a key locus influencing hepatic steatosis in a large-scale multi-ethnic meta-analysis. However, its comorbid mechanism linking the TyG index and MASLD remains uncharacterized, requiring experimental validation. Lysophosphatidic acid receptor 2 (LPAR2) mediates lysophosphatidic acid (LPA) signaling, promoting liver cell damage and inflammation. LPA signaling [particularly via lysophosphatidic acid receptor 1 (LPAR1)] drives fibrosis, and LPAR1 antagonists demonstrate anti-fibrotic potential. LPA actions also promote many features of liver cancer. Crucially, LPAR2 has negative PoPS scores in both TyG and MASLD, suggesting lower LPAR2 activity or expression may be protective. This aligns with LPAR2’s established pro-fibrotic/pro-inflammatory roles in liver disease. Suppressing its signaling could mitigate LPA-mediated damage, inflammation, and fibrosis throughout the TyG-MASLD spectrum. MAU2, a component of the chromosomal condensin complex, has been associated with hepatic steatosis in normal-weight individuals through its genetic variants. Studies have demonstrated that gene–gene and gene–environment interactions within the NCAN-TM6SF2-CILP2-PBX4-SUGP1-MAU2 locus significantly influence hyperlipidemia. MAU2 has negative PoPS scores in both TyG (−0.4518) and MASLD (−0.2123), suggesting reduced function may promote lipid problems and liver fat. This may occur via indirect effects on chromatin organization and gene expression at its locus, impacting lipid-regulating genes. NDUFA13, a nuclear-encoded subunit of mitochondrial complex I (NADH: ubiquinone oxidoreductase), maintains essential oxidative phosphorylation (OXPHOS) function. When NDUFA13 is impaired due to genetic mutations, posttranslational modifications, or oxidative damage, it induces mitochondrial dysfunction. In hepatocytes, this triggers a vicious cycle: energy deficiency and oxidative stress disrupt insulin signaling, exacerbating hepatic insulin resistance. These pathological changes contribute to the development of metabolic disorders. The TM6SF2 rs58542926 mutation impairs liver with very-low-density lipoprotein (VLDL) secretion. This causes fat buildup in liver cells (steatosis) and abnormal blood lipids, promoting fatty liver disease and heart risks. PoPS analysis shows a striking contrast: TM6SF2 has a strong positive score in TyG (7.2491) but a weak negative score in MASLD (−0.0416). This highlights tissue-specific effects. In TyG, the mutation directly raises blood TGs (positive score) by blocking VLDL secretion, driving the TyG index. However, in the liver, this mutation primarily causes severe steatosis. Its direct association with advanced MASLD complications appears attenuated, potentially modified by other factors, resulting in a weak negative correlation. MEF2B shows opposing associations (TyG: −0.7089; MASLD: +0.2001). Reduced activity in peripheral tissues during insulin resistance impairs metabolic adaptation, whereas elevated activity in steatosis drives pro-fibrogenic programs in hepatic stellate cells (HSCs) – consistent with its role in NASH-related macrophage chemotaxis. NRBP1 shifts from weak positive (TyG: +0.0609) to negative (MASLD: −0.3126), suggesting dual functionality: mild support for growth factor signaling in early dysmetabolismfailure of ubiquitin-dependent constraint on hepatic inflammation. ZNF513 shows discordant weak effects (TyG: +0.3129; MASLD: −0.1206), suggesting context-dependent transcriptional regulation. It may modulate lipid metabolism genes in early disease but dysregulate stress-response pathways in chronic steatosis. [] 57 [] 58 [] 59 [] 59 [] 60 [] 61 [] 59 [] 62 [] 63 [] 64 [] 65 [] 66 [] 67 [,] 68 69 [] 70 [] 70 [] 71 [] 72 [] 73 [] 74 [] 75 [,] 76 77 versus
The apolipoproteins APOA1, APOB, and APOC4 are mechanistically central to the comorbidity between the TyG index and MASLD. APOA1, APOB, and APOC4 are core members of the apolipoprotein family that play crucial roles in regulating lipid metabolism and insulin resistance. APOA1 serves as the primary structural protein of high-density lipoprotein. Reduced APOA1 expression impairs reverse cholesterol transport, directly promoting hepatic lipid accumulation and exacerbating insulin resistance– core pathways in MASLD. APOB is the key structural component of atherogenic lipoproteins (low-density lipoprotein and VLDL). Its overexpression drives abnormal VLDL secretion from hepatocytes, elevating circulating TGs, which worsens hepatic steatosis and insulin resistance. APOC4, a member of the apolipoprotein C family, contributes to dysregulation by impairing plasma TG clearance efficiency, further promoting intrahepatic lipid deposition. These genes maintain nonredundant roles in lipid homeostasis. Deleterious mutations cause severe dyslipidemias, driving strong purifying selection against nonfunctional variants. APOA1 has conserved domains essential for lipid binding and lecithin-cholesterol acyltransferase activation. APOB contains indispensable regions for lipoprotein assembly under intense selective constraint. APOC4 is in a conserved, tightly regulated apolipoprotein gene cluster. Thus, our sc-ST analysis in embryonic mouse tissue identified APOA1, APOB, and APOC4 as conserved nodes, though not all 12 human comorbidity genes. The other genes likely missed due to species differences in regulation, polygenic traits, or embryonic model limitations for human comorbidities. Genes affecting risk through subtle regulation or gene–environment interactions often diverge more evolutionarily. The conserved dysfunction in these apolipoproteins confirms the value of mouse models for core TyG-MASLD pathways. Key mechanisms – impaired fatty acid β-oxidation and faulty VLDL secretion driven by APOA1/APOB/APOC4, are recapitulated. Thus, despite species differences in some susceptibility genes, mouse models reliably reveal central lipid transport defects in TyG-linked MASLD, especially within these conserved pathways. [,] 78 79 [,] 78 79 [–] 80 82 [,] 81 82 [–] 83 85 [] 78 [] 80 [] 86 [] 86
Our multi-omics framework identified tissue-specific signatures for the TyG index and MASLD. TyG’s strong adipose association reflects its function as an insulin resistance marker (based on TGs/GLU), influenced by adipose processes including lipolysis, adipokine secretion, and inflammation. Vascular enrichment confirms the link between TyG, endothelial dysfunction, and atherosclerosis. Clinical studies show that TyG predicts cardiovascular events and arterial stiffness. Gastrointestinal enrichment in MASLD highlights the critical gut-liver axis. Dysbiosis, increased gut permeability, and bacterial translocation drive hepatic inflammation and steatosis progression, as shown by clinical and translational research. MASLD links to kidney disease, indicating shared metabolic risks and chronic kidney disease connections via systemic inflammation, insulin resistance, and RAAS activation. Embryonic sc-ST mapping localized signals to liver, connective tissue, and adrenal gland. The organ-wide Cauchy test was significant only in the adrenal gland, but liver/connective tissue spatial hotspots align with core MASLD pathology: hepatocyte lipid accumulation and HSC activation driving fibrosis. The adrenal gland’s significance in both sc-ST density and Cauchy tests suggests an underappreciated endocrine role. Dysregulated glucocorticoid/catecholamine stress responses may drive metabolic-inflammatory changes, requiring clinical studies on adrenal-MASLD severity links. [] 17 [] 87 [,] 88 89 [] 4
Our study has limitations requiring consideration. The MR analysis was restricted to European-ancestry populations, limiting generalizability. TWAS utilized GTEx v8 postmortem expression data, potentially introducing tissue-degradation biases unrelated tometabolic states. TyG data were filtered to 192 pre-selected SNPs rather than raw measurements, which may yield incomplete TWAS results – though this constraint originates from the source data. Although identified genes represent therapeutic targets, their mechanistic roles in endothelial dysfunction require validation through single-cell sequencing, animal models, and multi-ethnic replication to guide clinical translation. Collectively, these limitations underscore the need for broader population sampling and functional genomics to advance precision management of TyG-MASLD. in vivo
Conclusion
This study establishes the TyG index as a causal driver of MASLD. We identified 12 comorbid genes (e.g., TM6SF2 and GCKR) and conserved apolipoproteins (APOA1, APOB, and APOC4) that link hepatic lipid dysregulation to systemic cardiometabolic disturbances by orchestrating lipid trafficking, insulin signaling, and hepatic stress responses. These molecular hubs connect steatosis progression to cardiovascular risks, providing key insights into the unified etiology of CLMH disorders. Validating TyG as a causal biomarker and pinpointing these conserved pathways advances the CLMH field and highlights them as therapeutic targets. Clinically, integrating TyG-associated genetic variants with established noninvasive fibrosis biomarkers could enhance early identification of high-risk MASLD patients requiring intensified monitoring or confirmatory imaging. Future research should prioritize functional validation of these targets and explore combinatorial interventions targeting the TyG-MASLD-CVD axis to mitigate dual hepatic-cardiovascular risks.