Integrative functional genomics reveals transcriptional regulatory function of risk alleles for metabolic liver disease

Integrative functional genomics reveals transcriptional regulatory function of risk alleles for metabolic liver disease | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Integrative functional genomics reveals transcriptional regulatory function of risk alleles for metabolic liver disease Biying Zhu, Na He, Yang Xiao, Bin Chen, Chen Li, Ravi Mandla, and 12 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6984670/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Genome-wide association studies (GWAS) have identified nearly 100 loci associated with metabolic dysfunction-associated steatotic liver disease (MASLD), but the molecular functions of these variant alleles remain elusive, particularly when they occur in non-coding regions. Here we profiled the chromatin accessibility landscape of liver nuclei from MASLD individuals, and demonstrated these accessible genomic sites were bound by cell type-specific transcription factors (TFs) and enriched for MASLD risk variants, highlighting lineage- and disease state-specific regulation. Using a massively parallel reporter assay (MPRA), we identified hundreds of differential activity variants (DAVs) that operate in a cell type-specific manner or in a stimulus-dependent context by disrupting liver pathogenesis-associated transcriptional regulatory network. Integrative analyses combining liver eQTLs, chromatin looping, and single-cell CRISPRi screening linked these DAVs to functional target genes. Notably, we demonstrated that DAVs located near SLC22A3 and core regulators of triglyceride metabolism ( APOA5 , ANGPTL3 , and LPL ) loci modulate their gene expression and contribute to altered lipid metabolism and hepatic stellate cell activation. Furthermore, these DAVs exhibit predictive power in distinguishing MASLD disease risk. Together, these multimodal integration analyses provide insights into the regulatory mechanisms of MASLD progression driven by noncoding genetic risks. Biological sciences/Genetics/Functional genomics Health sciences/Diseases/Metabolic disorders Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Metabolic dysfunction-associated steatotic liver disease (MASLD), the most common chronic liver condition, encompasses a spectrum from metabolic dysfunction-associated steatotic liver (MASL), characterized by simple lipid accumulation, to advanced metabolic dysfunction-associated steatohepatitis (MASH), marked by progressive inflammation and fibrosis 1 . Given the limited therapeutic options for MASLD, there is an urgent need to elucidate its underlying causal mechanisms. MASLD arises from the complexed interplay between environmental factors and genetic predisposition, with heritability estimates ranging from 20% to 70% 2 . Large-scale genome-wide association studies (GWAS) have identified nearly 100 loci associated with MASLD risk, but most variants lie in noncoding regions with unknown functions 3 – 7 . Furthermore, high linkage disequilibrium (LD) and genetic pleiotropy complicate the identification of causal variants and target genes. Numerous computational methodologies, including fine-mapping, expression quantitative trait loci (eQTL) analysis, chromatin accessibility quantitative trait loci (caQTL) analysis, and deltaSVM algorithms, have been employed to identify causal functional variants, but they have demonstrated limited predictive power 8 – 10 . Noncoding genetic variants can influence gene expression by altering chromatin accessibility and transcription factor binding at gene regulatory elements (GREs), such as enhancers 11 , 12 . Massively parallel reporter assays (MPRA) offer a scalable experimental approach to test the enhancer activity of risk alleles in disease-relevant cell types, organoids, or organisms under various environmental conditions 13 – 15 . Since GREs are often cell-type-specific, it is crucial to understand the role of disease-associated variants within relevant cell types. While many studies using single-cell/nuclei RNA sequencing (scRNA-seq or snRNA-seq) have shown that MASLD pathophysiology involves multiple liver cell types 16 , 17 , the chromatin landscape heterogeneity in MASLD patient livers has been rarely explored. This understanding is essential for interpreting the molecular mechanisms underlying noncoding MASLD risk variants. Here, we profiled the chromatin accessibility landscape of liver nuclei from MASLD patients and found that accessible regions were enriched for MASLD risk variants and bound by cell type-specific transcription factors. MPRA screening identified hundreds of differential activity variants (DAVs) with context-specific regulatory activity that perturb transcriptional networks associated with liver pathogenesis. Integration with liver eQTL data, chromatin interaction maps, and single-cell CRISPRi screening enabled the assignment of target genes to these regulatory variants. Notably, we validated key DAVs that modulate lipid metabolism and drive hepatic stellate cell activation, the hallmarks of MASLD pathology. Importantly, these DAVs exhibit predictive power in distinguishing MASLD disease risk. Thus, our multimodal analyses reveal how noncoding genetic variation drives MASLD progression through disruption of cell type- and state-specific regulatory programs. Results Cell type-specific regulation of chromatin accessibility by MASLD risk variants To delineate the regulatory program of MASLD-associated genetic noncoding variants, we profiled the chromatin landscape of MASLD patient liver. We performed single-nucleus ATAC sequencing (snATAC-seq) on liver nuclei from 12 individuals (6 Normal and 6 MASLD) ( Fig. 1 A ) , and identified 69,595 high-quality nuclei and classified them into 15 distinct cell clusters ( Fig. 1 B and Extended Data Fig. 1 A- 1 K ) . Each cluster displayed a unique chromatin accessibility landscape and TF motif enrichment signature, reflecting robust cell identity-specific regulatory programs (Extended Data Fig. 2 A- 2 C ) . Our cell type annotations were cross-validated using a recently published human liver snATAC reference dataset (Extended Data Fig. 2 D- 2 E ) 18 . Furthermore, by inferring the gene scores of marker genes of each cluster (Extended Data Fig. 2 F ) , we identified clusters representing hepatocytes ( ALB ), cholangiocytes ( KRT7 ), Kupffer cells ( CD163 ), stellate cells ( NGF ), endothelial cells ( FLT1 ), and T/NK/B cells ( CD8A ), with each cell cluster expressing a unique transcriptional program ( Fig. 1 B and Extended Data Fig. 2 G ) . We observed highly specific correlations between marker genes defined by snATAC-seq and previous snRNA-seq from human MASLD liver (Extended Data Fig. 2 H ) 17 . As expected, MASLD liver samples consisted of higher percentages of cholangiocytes and stellate cells, and decreased hepatocyte and endothelial cells, compared to normal livers ( Fig. 1 C and Extended Data Fig. 1 H- 1 J ) 19 . We next sought to test whether MASLD-associated variants were enriched in liver regulatory elements. Stratified linkage disequilibrium score regression (LDSC) analysis demonstrated enrichment of MASLD GWAS variants in hepatocyte and stellate cell chromatin peaks, while plasma triglyceride trait-associated variants were enriched in accessible genomic regions of hepatocytes and Kupffer cells (Extended Data Fig. 3 A- 3 B ) . In contrast, no enrichment was observed for variants associated with body mass index (BMI) traits. To further investigate the heterogeneity of genetic association enrichment, we applied SCAVENGE, an algorithm we developed that integrates scRNA-seq and GWAS data to identify disease-associated genetic variants at single-cell resolution 20 . Our analysis revealed that liver fat trait-associated variants were specifically enriched in hepatocyte peaks, whereas MASLD-associated variants displayed a broader pattern of enrichment across various cell types, including hepatocytes, stellate cells, and Kupffer cells ( Fig. 1 D- 1 E ) , suggesting the genetic involvement of non-parenchymal cells in addition to hepatocytes in MASLD. We then focused on stellate cells, the primary effector cells in MASLD progression 21 , and identified three sub-clusters with varying chromatin accessibility. Based on the cluster composition of cell origin, we termed these three stellate cell clusters as S-MASLD, S-Normal, and S-Common ( Fig. 1 F- 1 G ) . The proportion of S-MASLD cells was higher in MASLD liver compared to normal liver. Genomic region enrichment analysis of S-MASLD-specific peaks revealed pathways related to extracellular matrix organization and inflammatory responses (Extended Data Fig. 3 C- 3 D ) . Strikingly, MASLD genetic risk variants were differentially enriched among stellate subtypes, with S-MASLD cells exhibiting the highest SCAVENGE trait relevance scores (TRS) ( Fig. 1 H ) . Pseudotime trajectory analysis, validated by two independent algorithms, revealed a continuous activation process across stellate cell states, during which MASLD risk variants became progressively enriched ( Fig. 1 I- 1 K and Extended Data Fig. 3 E- 3 F ) . Beyond stellate cells, we observed substantial chromatin heterogeneity among hepatocyte subpopulations. Disease-enriched hepatocyte states showed preferential localization of MASLD-associated risk variants ( Fig. 1 L- 1 O and Extended Data Fig. 3 G- 3 J ) . A similar trend was observed in Kupffer cells, where disease-associated Kupffer cells exhibited notably higher SCAVENGE TRS scores (Extended Data Fig. 4 A- 4 F ) . Collectively, these findings indicate that MASLD risk variants exert cell type- and cell state-specific effects on the hepatic regulatory landscape to drive disease progression. Given our ability to simultaneously map GWAS variant enrichments using SCAVENGE and TF motif activities using chromVAR at single-cell resolution, we next performed integrative analyses to gain mechanistic insight into how noncoding risk variants might influence transcriptional regulation in MASLD 10 . We observed a positive correlation between MASLD enrichment TRS scores and the motif activities of TFs related to metabolism and fibrosis, such as SNAI2, across all profiled nuclei ( Fig. 1 P ) 22 . A similar trend was seen when the analysis was performed in hepatocytes alone, where CEBPA motif activities were positively correlated with MASLD trait TRS scores (Extended Data Fig. 4 G ) . Additionally, integrative analysis of cirrhosis, a more advanced stage of MASLD, revealed the strongest positive correlations with TFs like CEBPA, HNF4A, and NR2F6, and negative correlations with TFs such as GABPA ( Fig. 1 Q and Extended Data Fig. 4 H- 4 I ) , which is involved in oxidative phosphorylation 23 . To further dissect the regulatory architecture of MASLD progression, we integrated snRNA-seq and snATAC-seq data to construct hepatocyte-specific transcriptional regulatory networks using Pando 24 . We revealed extensive disease-dependent transcriptional regulatory network rewiring, characterized by altered expression and motif enrichment of key TFs implicated in MASLD pathogenesis, including FOXO3, MLXIPL, PROX1, PPAR, HNF1B/4A, and TEAD1 (Extended Data Fig. 4 J-K ) . Identify and characterize the candidate functional variants using MPRA To systematically identify functional noncoding variants associated with MASLD, we designed an MPRA library that includes both alleles of 5,369 noncoding variants in strong linkage disequilibrium (r² >0.8) with genome-wide significant as well as subthreshold significant variants from recent GWASs (Extended Data Fig. 5 A and Table S1 ) 4 , 6 , 7 , 25 – 30 . Notably, we included subthreshold variants in the MPRA library based on their potential to modulate enhancer function and contribute to MASLD risk, despite not surpassing conventional significance thresholds for GWAS 31 , 32 . Interestingly, these MPRA candidate variants were predominantly enriched in regions of open chromatin within primary liver cell types, such as hepatocytes and immune cells (Extended Data Fig. 5 B- 5 D ) . The MPRA was conducted in disease-relevant hepatic cell lines, specifically the hepatocyte cell line HepG2 and the stellate cell line LX-2, both with and without environmental stimuli (fatty acid and TGFβ) to simulate environmental risk factors associated with MASLD ( Fig. 2 A ) . The MPRA library was synthesized with a median library complexity of 714 barcodes per allele and included 40 positive regulatory elements predicted to be active in hepatocytes (Extended Data Fig. 5 E ) . The positive regulatory elements demonstrated elevated transcriptional activity, whereas the negative regulatory elements exhibited minimal activity in HepG2 cells, validating the robustness of the MPRA approach (Extended Data Fig. 5 F- 5 G ) . MPRA was performed with four replicates for each condition, and exhibited high reproducibility, as exemplified in HepG2 (Extended Data Fig. 5 H ) . Overall, 2430 regulatory elements were transcriptionally active for at least one allele in HepG2 and enriched for metabolic TFs such as HNF4, CEBPD and TRβ ( Fig. 2 B and Extended Data Fig. 6 A ) . We subsequently assessed allele-specific activity within these active elements, identifying 365 differential activity variants (DAVs) [FDR < 0.01] distributed across 21 chromosomes ( Fig. 2 C, Extended Data Fig. 6 B- 6 C and Table S2 ) . To validate the sensitivity and specificity of the DAVs identified by MPRA, we randomly selected 10 DAVs with high allele-specific activity and examined their activities individually. Notably, eight out of ten DAVs demonstrated the expected allele-specific activity in HepG2 cells ( Fig. 2 D ) , confirming the reliability of the MPRA. We next characterized the properties of the DAVs through integrative analysis, incorporating chromatin accessibility information identified by snATAC-seq, liver eQTL data, liver or HepG2 chromatin interaction data, TF motif disruption, and liver chromatin functional annotations 33 , 34 . Nearly 90% of DAVs were located within active genomic regions, and 297 out of 365 DAVs overlapped with two or more functional annotations ( Fig. 2 E and Extended Data Fig. 6 D- 6 E ) . Of particular interest were 87 DAVs that were located in chromatin-accessible regions across various cell types, with higher enrichment observed in open chromatin regions of hepatocytes, stellate cells, and cholangiocytes ( Fig. 2 F and Extended Data Fig. 6 F- 6 H ) . For example, several DAVs located in the GLDC and MLXIP loci reside within hepatocyte-specific chromatin peaks and were associated with strong enhancer signals (Extended Data Fig. 7 A- 7 B ) . Notably, 274 DAVs altered the motif activity of specific TFs. We observed a positive correlation between TF binding disruption scores and MPRA effect sizes for the DAVs ( Fig. 2 G ) . Specifically, HNF4G, SREBF1/2, and ETS (Erythroblast Transformation-Specific) family binding sites were enriched for DAVs, as determined using two different algorithms, SNP2TFBS and MotifbreakR ( Fig. 2 H and Extended Data Fig. 6 I ) . These results are consistent with the established roles of HNF4 and SREBF1/2 in liver metabolism and underscore the involvement of ETS family transcription factors in MASLD progression 35 – 38 . Furthermore, we observed that motif activities of HNF4G and SREBF1/2 were higher in the MASLD liver compared to the normal liver, whereas the ETS family member GABPα binding activity was reduced in MASLD liver ( Fig. 2 I and Extended Data Fig. 8A-8E) . Together, these functional annotations allow us to prioritize potential regulatory variants and nominate their target genes with greater precision. DAVs exhibit cell type-specific regulatory function The progression of MASLD involves multiple liver cell types, including the activation of hepatic stellate cells. We hypothesized that MASLD risk variants modulate the activity of regulatory elements in a cell type-specific manner. MPRA analysis revealed substantial differences in regulatory element activity between human hepatocyte cells line HepG2 and stellate cell line LX-2, in line with the cell type-specific chromatin accessibility observed ( Fig. 3 A and Extended Data Fig. 9A) . Regulatory elements specific to HepG2 cells were enriched for motifs associated with PPARα and ERRγ, key nuclear receptors involved in liver metabolism. In contrast, LX-2-specific elements contained binding sites for fibrosis-related TFs, such as STAT6 ( Fig. 3 B and Extended Data Fig. 9B) . These cell type-specific regulatory element activities resulted in more than half of the DAVs exhibiting allele-specific activity exclusively in either HepG2 or LX-2 cells ( Fig. 3 C- 3 E and Extended Data Fig. 9C) . In total, we identified 464 DAVs in LX-2 cells, with the majority being liver eQTLs, located in epigenomic regions, or disrupting specific TF binding motifs (Extended Data Fig. 9D-9E and Table S3 ) . Notably, LX-2-specific DAVs disrupted the binding sites of STAT, TEAD, and TCF21 ( Fig. 3 F and Extended Data Fig. 9F) , all of which have been implicated in liver fibrosis 39 . For instance, the genetic variant rs6726823 near the CASP8 gene exhibited significant allele-specific activity in LX-2 but showed far less differential activity in HepG2 ( Fig. 3 G ) . Interestingly, the predicted target gene of rs6726823, CASP8 , has been functionally implicated in hepatic stellate cell homeostasis and liver fibrosis in MASLD mouse models 40 , 41 . Mechanistically, we found that rs6726823 resides in stellate-specific chromatin peaks and may regulate CASP8 expression by affecting the binding motif of IRF, a transcription factor family implicated in liver fibrosis ( Fig. 3 H- 3 I ) 42 . In contrast, the variant rs13150068 displayed allele-specific activity exclusively in HepG2 cells, with no notable differential activity in LX-2 ( Fig. 3 J ) . This variant is located in a hepatocyte-specific open chromatin region near HSD17B13 , which is a liver-enriched gene involved in lipid droplet metabolism and strongly associated with MASLD pathogenesis ( Fig. 3 K ) 43 . Motif analysis suggests that rs13150068 may exert its regulatory effect by disrupting the binding site of SREBP2, a master transcriptional regulator of lipid biosynthesis ( Fig. 3 L ) . These findings underscore the importance of selecting disease-relevant cell lines to accurately identify bona fide functional variants. Context-dependent activity of DAVs Since the progression of MASLD is shaped by the complex interplay between genetic predisposition and environmental factors 2 , we further delineated the functional variants under various environmental stimuli, including oleic acid and palmitic acid (PAOA), and TGFβ to mimic the metabolic stress in MASLD livers 44 . Hundreds of regulatory elements were induced in HepG2 under PAOA condition, with PAOA-induced elements being enriched for the binding motifs of metabolic nuclear receptors, ERRα and TRβ (Extended Data Fig. 10A-10C) . A total of 338 DAVs were identified in PAOA-treated hepatocytes, with most of them having at least two functional annotations ( Fig. 4 A- 4 B and Extended Data Fig. 10D-10E and Table S4 ) . Notably, one-third of the DAVs were stimulus-dependent, displaying variable allele-specific activity between control and PAOA conditions ( Fig. 4 C and Extended Data Fig. 10F) . Of note, these PAOA-specific DAVs appeared to preferentially affect the PPARE and RUNX1 motifs ( Fig. 4 D ) . In particular, rs12406250 exhibited higher allele-specific activity under PAOA conditions compared to control ( Fig. 4 E ) . This variant is located in an intronic open chromatin peak of the EPHA2 gene, whose expression and signaling are known to be altered during MASLD pathogenesis ( Fig. 4 F ) 45 . Moreover, the intensity of the peak containing rs12406250 was elevated in MASLD-hepatocytes relative to normal hepatocytes ( Fig. 4 F ) , suggesting the role of rs12406250 in regulating the chromatin accessibility in MASLD. We further identified 614 DAVs in LX-2 cells in response to TGFβ treatment, and these DAVs likely exert their regulatory effects by modulating chromatin accessibility and TF binding dynamics ( Fig. 4 G and Extended Data Fig. 10G-10I and Table S5 ) . Notably, most of these DAVs exhibited differential allele-specific activity compared to the basal state, suggesting that TGFβ alters the genomic binding landscape and reveals context-specific variant function ( Fig. 4 H- 4 J ) . For example, rs558702, a DAV located within a stellate cell-specific open chromatin region, exhibited enhancer activity exclusively upon TGFβ treatment, highlighting its context-specific regulatory role ( Fig. 4 K- 4 L ) . Together, these findings underscore how environmental cues such as PAOA and TGFβ exposure can unmask the functional consequences of noncoding variants and provide mechanistic insight into gene-environment interactions driving MASLD pathogenesis. Disease-associated transcriptional programs driven by DAVs To elucidate the transcriptional networks regulated by context-specific DAVs, we systematically analyzed the TFs disrupted by DAVs in HepG2 and LX-2 cells. Protein-protein interaction network analysis revealed that TFs perturbed by HepG2-DAVs formed a tightly connected transcriptional module centered around SREBP signaling and lipid metabolism, whereas LX-2-perturbed TFs were enriched in a network linked to TGFβ signaling ( Fig. 5 A- 5 D and Extended Data Fig. 11A-11B) . These results indicate that MASLD-associated DAVs converge on distinct, cell-type-specific transcriptional programs, with TFs functioning cooperatively within coordinated, disease-relevant regulatory circuits. To accurately identify the target genes of these DAVs, we integrated multiple orthogonal datasets, including public liver eQTL database, chromatin interaction data, and co-accessibility analyses from snATAC, as well as proximal gene information ( Fig. 5 E ) . This comprehensive approach enabled us to assign between 794 and 1,009 target genes for different set of DAVs, with 995 DAVs having at least one associated target gene ( Fig. 5 F- 5 G and Extended Data Fig. 12A-12F) . Among these, 338 DAVs overlapped with known liver eQTLs, and notably, 64% exhibited concordant effect directions between reporter activity and eQTL effect size ( Fig. 5 H and Extended Data Fig. 12G-12J) . The remaining 36% discordance likely reflects the cellular heterogeneity of liver tissue and the context-dependent nature of chromatin regulation. Interestingly, gene ontology analysis of these potential target genes revealed many biological processes associated with MASLD progression, including lipid metabolism, immune responses, and the TGFβ signaling pathway ( Fig. 5 I- 5 J ) . Moreover, transcriptional regulatory network analysis revealed that many DAV target genes are regulated by key MASLD-related TFs, including SREBF1, MLXIPL (ChREBP), ERG, and TCF7L2 (Extended Data Fig. 12K) . To validate the functional predictions, we deactivated the genomic regions spanning 8 randomly selected DAVs using CRIPSR interference (CRISPRi) approach and examined the expression levels of the nominated target genes. Over 90% of the potential target genes showed reduced expression levels, consistent with predictions ( Fig. 5 K and Extended Data Fig. 13A) . Conversely, when CRIPSR activation (CRISPRa) machinery was introduced into the cells, most target genes exhibited increased expression levels (Extended Data Fig. 13A-13B) . To systematically identify the target genes of MASLD-associated DAVs, we employed a CROP-seq approach that integrates scRNA-seq with CRISPR screening 46 . A lentiviral-based CROP-seq library was generated with high complexity and uniformity, comprising 40 sgRNAs targeting 20 genomic regions spanning MASLD-associated DAVs with strong allelic effects and well-annotated metabolic target genes (Extended Data Fig. 13C and Table S6 ) . scRNA-seq of CROP library-infected HepG2 cells revealed 67,700 high-quality cells and identified 18 distinct cell clusters (Extended Data Fig. 13D-13F) . We then associated the DAVs with the previously nominated target genes and identified 25 unique gene-sgRNA pairs, where the expression levels of the target genes were significantly lower in DAVs sgRNA-infected cells compared to all other sgRNA-infected cells ( Fig. 5 L ) . This analysis offers a robust strategy for narrowing down the pool of target genes associated with functional variants, providing mechanistic insights into MASLD progression. Functional characterization of DAVs modulating APOA5, ANGPTL3, and LPL in triglyceride metabolism With the identification of potential functional DAVs and their annotated target genes, we next investigated how these DAVs influence lipid metabolism, a key aspect of MASLD, in the liver. APOA5, ANGPTL3/8, and LPL form a tightly regulated network that plays a central role in triglyceride metabolism and systemic lipid homeostasis 47 . We then focused on two DAVs (rs1748197 and rs10889356) that are near the ANGPTL3 locus ( Fig. 6 A ) , a well-known inhibitor of LPL activity 48 , and have been previously identified as risk variants associated with MASLD 7 . Although neutralizing antibodies targeting ANGPTL3 have been approved to reduce lipoprotein levels in individuals with homozygous familial hypercholesterolemia, the inhibition of hepatic ANGPTL3 using antisense oligonucleotide (ASO) drug Vupanorsen unexpectedly led to liver lipid accumulation, ultimately resulting in the termination of the clinical trial 49 , 50 . We discovered that DAV rs1748197 is located in the promoter region of the ANGPTL3 gene, and confirmed that rs1748197 acts as an eQTL variant for ANGPTL3 ( Fig. 6 A- 6 B ) . Deletion of a small genomic region containing rs1748197 or deactivation of this genomic region using CRISPRi reduced the expression of ANGPTL3 , while genomic activation using CRISPRa enhanced the expression of ANGPTL3 ( Fig. 6 D- 6 F and Extended Data Fig. 14A) . Mechanistically, rs1748197 is predicted to alter the binding motif of MYC, a transcription factor implicated in the pathogenesis of alcoholic liver disease (Extended Data Fig. 14B) 51 . Meanwhile, another DAV rs10889356 resides in an enhancer region that interacts with the ANGPTL3 promoter ( Fig. 6 A ) . Public liver eQTL data reveals a strong association between rs10889356 and ANGPTL3 expression, with livers carrying G allele having higher ANGPTL3 expression ( Fig. 6 C ) . Indeed, inhibition or activation of this enhancer region reduced or increased ANGPTL3 expression, respectively (Extended Data Fig. 14C) . Motif analysis further revealed that rs10889356 perturbs a GABPA binding motif, and reporter assays confirmed that the G allele promotes enhancer activity, likely through increased GABPA recruitment (Extended Data Fig. 14D-14E) . Functionally, deletion of the small genomic regions containing rs1748197 and rs10889356 individually in HepG2 cells resulted in greater lipid accumulation compared to wild-type HepG2 cells under both basal and PAOA-treated conditions ( Fig. 6 G- 6 H and Extended Data Fig. 14F-14G) . These findings were further confirmed in these DAVs-deleted hepatocyte spheroids treated with PAOA (Extended Data Fig. 14H) . These data suggest that MASLD-associated DAVs rs1748197 and rs10889356 modulate the expression of ANGPTL3 , which negatively regulates hepatic lipid metabolism. We further identified rs10790167 as a HepG2-specific DAV with robust allele-specific activity and no significant regulatory effect in LX-2 cells ( Fig. 6 I ) . This variant is located in a hepatocyte-specific open chromatin region and establishes an enhancer-promoter (E-P) interaction with the promoter of APOA5 , which is a known activator of LPL activity and closely linked to MASLD progression 52 . Motif analysis revealed that rs10790167 disrupted the binding motifs of several nuclear receptors, including PPAR and RXR (Extended Data Fig. 14I) . Deletion of rs10790167 significantly reduced the expression of APOA5 and led to marked accumulation of lipid droplets in HepG2 cells, suggesting a causal role for this variant in dysregulated lipid metabolism ( Fig. 6 J- 6 L ) . In contrast, the genetic variant rs1441754 near the lipoprotein lipase ( LPL ) gene, previously identified as a cholesterol-associated variant, exhibited significant allele-specific activity in LX-2 but showed far less differential activity in HepG2 ( Fig. 6 M- 6 N ) 53 . Luciferase reporter assays confirmed this cell type-specific enhancer activity, with significant allele-specific effects observed exclusively in LX-2 cells, but not in HepG2 or Huh7 hepatocyte cell lines (Extended Data Fig. 14J) . Interestingly, the predicted target gene of rs1441754, LPL , is upregulated in hepatic stellate cells and exacerbates liver fibrosis in MASLD in mouse models 54 . Mechanistically, we found that rs1441754 resides in stellate-specific chromatin peaks and may regulate LPL expression by affecting the binding motif of RUNX (Extended Data Fig. 14K) . Strikingly, deletion of the rs1441754-containing region in LX-2 resulted in a significant increase in LPL expression, suggesting its role as a negative regulatory element in stellate cells ( Fig. 6 O ) . Functionally, rs1441754 deletion also resulted in the upregulation of multiple fibrosis-associated genes in LX-2 cells ( Fig. 6 O- 6 P ) . Together, these findings highlight the integrative roles of context-specific DAVs in modulating key regulators of triglyceride metabolism including APOA5, ANGPTL3, and LPL, and their downstream contributions to altered lipid homeostasis and hepatic stellate cell activation in MASLD ( Fig. 6 Q ) . In addition to the role of rs1441754 in stellate cell activation, we identified another regulatory variant, rs66477683, which modulates MERTK expression (Extended Data Fig. 15A-15B) . Given prior reports linking MERTK to liver fibrosis 55 , our findings suggest that rs66477683 may represent a fibrosis-driving variant acting through transcriptional control of MERTK (Extended Data Fig. 15C-15D) . Functional linking of rs474513 to SLC22A3 and hepatic lipid homeostasis Notably, DAV rs474513, previously linked to obesity and type 2 diabetes 56 , is located in hepatocyte-specific open chromatin regions near the promoter of the SLC22A3 gene. Chromatin interaction data also revealed a strong enhancer-promoter interaction between the genomic region containing rs474513 and the LPA gene, suggesting the regulatory potential of rs474513 on SLC22A3 and LPA ( Fig. 7 A ) . Interestingly, public liver eQTL data showed that the genotype of rs474513 is closely associated with the expression levels of both SLC22A3 and LPA , but in opposite directions ( Fig. 7 B ) . Livers carrying the alternative allele G exhibited higher expression of SLC22A3 and lower expression of LPA compared to those carrying the reference allele A. Mechanistically, rs474513 modulates the binding motif for RUNX, a transcription factor known to be involved in MASLD progression 57 . The agreement score with the RUNX consensus motif was 997-fold greater for the G allele than for the A allele ( Fig. 7 C ) . Consistently, overexpression of RUNX1/2 increased the activity of reporters carrying the G allele compared to those with the A allele ( Fig. 7 D ) . In contrast, inhibition of RUNX2 activity using small molecule CADD522 reduced SLC22A3 expression, but had no effect in HepG2 cell where a small genomic region containing rs474513 was deleted (Extended Data Fig. 16A-16B) . Moreover, deletion of the genomic region containing rs474513 led to decreased SLC22A3 expression and increased LPA expression, in line with the eQTL data ( Fig. 7 E ) . Similar gene expression patterns were observed when the genomic region was inactivated using CRISPRi technology (Extended Data Fig. 16C) . Importantly, the G > A-edited heterozygous HepG2 cells also exhibited decreased SLC22A3 expression ( Fig. 7 F and Extended Data Fig. 16D) . To examine the role of rs474513 in lipid metabolism associated with MASLD, we treated the rs474513-deleted or gene-edited HepG2 with PAOA to model steatosis. Interestingly, we found that lipid accumulation was increased in rs474513-deleted and G > A-edited HepG2 cells treated with PAOA ( Fig. 7 G- 7 J ) . Given the established role of RUNX in mediating the transcriptional activity of rs474513, we next treated HepG2 with the RUNX1/2 inhibitors Ro5-3335 and CADD522. Pharmacologic inhibition of RUNX signaling led to reduced SLC22A3 expression and further exacerbated lipid accumulation, mirroring the effects of rs474513 deletion ( Fig. 7 K and Extended Data Fig. 16E-16F) . Importantly, ectopic overexpression of SLC22A3 rescued lipid accumulation phenotype ( Fig. 7 K and Extended Data Fig. 16E-16F) , confirming that SLC22A3 acts downstream of the rs474513-RUNX regulatory axis to maintain hepatic lipid homeostasis. To further investigate the physiological function of rs474513 in MASLD-relevant organoid model 58 , we deleted a small genomic region containing rs474513 in human induced pluripotent stem cell (hiPSC) line and observed a reduction in SLC22A3 expression (Extended Data Fig. 16G-16H) , consistent with the above finding. Importantly, the deletion of rs474513 did not impair the efficiency of hepatic differentiation in the hiPSC line (Extended Data Fig. 16I) . We then differentiated both the wild-type and rs474513-depleted hiPSC lines into human liver organoids (HLOs) and exposed them to free fatty acids to model the steatosis process 58 . Interestingly, the rs474513-depleted HLOs exhibited higher lipid accumulation compared to wild-type HLOs, along with increased expression of PLIN2 , a gene involved in lipid storage and metabolism ( Fig. 7 L- 7 M ) . We also observed elevated levels of CD68 and CK7 in rs474513-depleted HLOs ( Fig. 7 L- 7 M ) , indicating an enhanced inflammatory response and ductal reaction. Furthermore, the higher ACTA2 and lower ALBUMIN expression implied the potential risk of fibrosis. Functional DAVs improve MASLD risk prediction Polygenic risk scores (PRS) aggregate the effects of multiple genetic variants to estimate an individual's likelihood of developing a disease 59 . To assess whether hepatic DAVs can effectively predict MASLD risk, we compared the performance of the PRS constructed from 114 LD-independent MPRA-prioritized variants (DAVs) (Table S7 ) to the control PRS (MPRA-Ctrl) constructed from randomly selected, matched set of MPRA total variants using data from the Million Veteran Program (MVP) 60 . Our analysis revealed that hepatic DAVs PRS outperformed the control PRS in predicting various MASLD-related traits, including MASLD defined by chronic elevation of liver enzyme alanine aminotransferase (MASLD-ALT), MASLD defined by international classification of disease (ICD) codes (MASLD.ICD), ALT levels, and fibrosis index FIB4 progression ( Fig. 7 N ) . Notably, the individuals with DAV PRS scores in the top 10% exhibited a 1.6-fold higher risk of developing MASLD compared to the ones in the lowest 10%, which is better than the performance of MPRA-Ctrl PRS, and comparable to the PRS built from nine well-established missense variants contributing to MASLD (PRS9), including the rs738409 (I148M) variant of the gene PNPLA3 and rs58542926 (E167K) of the gene TM6SF2 ( Fig. 7 O and Table S8 ) 4 . We further benchmarked the DAV-PRS in the All of Us Research Program against previously published MASLD PRS models and found that, while our MPRA-prioritized PRS performed similarly to PRS9, its predictive power was modestly lower than PGS002732, the best-performing published model (Extended Data Fig. 16J) 4 . These data underscore the power of our approach in identifying functional variants linked to MASLD, which can also be extended to other complex diseases. Discussion Fine-mapping causal variants for diseases or complex traits continues to be a major hurdle in the field of genetics 61 . Numerous computational algorithms have been developed, employing various statistical and computational strategies to fine-map genetic variants. For example, FINEMAP and DAP utilize Bayesian frameworks to estimate posterior probabilities of causality; PAINTOR and fGWAS incorporate functional annotations to prioritize variants; Coloc and eCAVIAR identify shared causal variants between complex traits 62 . Additionally, deltaSVM enhances fine-mapping by providing functional predictions for non-coding variants that disrupt transcription factor binding using machine learning techniques 63 . However, computational prediction algorithms for prioritizing functional variants are often not concordant among themselves and are frequently inconsistent with experimental validation 9 . Here, by integration of snATAC, high-throughput MPRA, and CROP-seq data, we identified functional causal variants and revealed key disease-associated cell types or cell states and elucidated the transcriptional programs influenced by these functional variants. Remarkably, functional variants identified in this way regulate hepatic lipid metabolism and stellate cells activation and effectively predict MASLD risk. To our knowledge, this is the first study to comprehensively profile the heterogeneity of chromatin landscapes in individuals at different stages of MASLD and thus provides a high-quality human liver regulome resource. Integrating GWAS variants with single-cell chromatin landscape data revealed that MASLD risk variants are enriched in disease-specific hepatocyte and stellate cell states. Moreover, by combining motif activity and risk variant enrichment analyses at the single-cell level, we demonstrated how these MASLD risk variants regulate gene expression. Furthermore, the single-cell chromatin accessibility data enabled us to nominate target genes of risk variants using co-accessibility analysis. Notably, snATAC-seq studies using large cohorts of MASLD liver samples will enhance the statistical power to pinpoint the functional variants by performing chromatin accessibility QTL (caQTL) or allele-imbalance analyses 64 . Importantly, our MPRAs were conducted using two disease-relevant cell lines under MASLD-associated environmental conditions, demonstrating that these DAVs exhibit cell type-specific and stimulus-dependent activity. This highlights the importance of selecting appropriate cell lines and cellular contexts for identifying functional SNPs. Notably, only 64% of DAVs exhibited concordant effect directions between reporter activity and eQTL effect size, while the remaining 36% showed discordance, likely reflecting the cellular heterogeneity of liver tissue and the context-dependent nature of chromatin environments. Future implementation of this approach in human liver organoids or in vivo models may provide more physiologically relevant insights 14 . Moreover, using various CRISPR approaches, we identified MASLD-associated transcriptional programs. We observed that rs474513 affects the expression of SLC22A3 and LPA in opposite directions, demonstrating how a single variant can disparately regulate multiple genes 65 . Conversely, both rs1748197 and rs10889356 regulate the expression of ANGPTL3 , suggesting synergistic effects of these genetic variants. Remarkably, we identified several MASLD-associated DAVs that regulate the core regulatory network of LPL activity, including APOA5 (an LPL activator), ANGPTL3 (an LPL inhibitor), and LPL itself. These findings suggest that MASLD development is possibly mediated by local paracrine effects on LPL activity in the liver. Importantly, using physiologically relevant human liver organoids and gene-editing tools, we demonstrated the causal roles of these genetic variants in lipid metabolism associated with MASLD. In sum, our study combined high-throughput MPRA, snATAC-seq, and CROP-seq to identify functional causal variants and their target genes in MASLD-related cell types. This broadens our understanding of genetic mechanisms underlying MASLD risk, and will be a valuable guide to prediction and treatment of MASLD and related disorders. Moreover, this approach may be employed to better understand other complex human diseases. Declarations Acknowledgements We thank all other members of the Hu and Lazar labs for their invaluable help and discussion. We would also like to thank the Liver Tissue Cell Distribution System and SEKISUI XenoTech for providing the MASLD and healthy liver samples. We thank Shixiang Wang from Dr. Yun Shen Chan lab for the help with the hepatic organoid differentiation. We thank Dr. Yajie Zhao for the discussion on PRS analysis. This PRS analysis is based on data from the Million Veteran Program (MVP), Office of Research and Development, Veterans Health Administration with grateful support by award MVP000 and Veterans who participated in MVP. This publication does not represent the views of the Department of Veterans Affairs, the US Food and Drug Administration or the US Government. This work was supported by the National Key R&D Program of China (2024YFF0509000 to W.H.), the National Natural Science Foundation of China (82270868 to W.H.), the R&D Program of Guangzhou National Laboratory (GZNL2023A02004, GZNL2023A03004, and GZNL2023A02002 to W.H.), the Pearl River Talent Recruitment Program (2021QN02Y963 to W.H.), National Institutes of Health grant (R01 DK125573, R01DK49780, and P30 DK019525 to M.A.L., R01 DK134575 to V.M.), the JPB Foundation (to M.A.L.), the Cox Medical Institute (to M.A.L.), VA Merit Funding BX003362 from VA Office of R&D (K.M.C.), and VA Informatics and Computing Infrastructure (VINCI) VA HSR RES 130457 (J.A.L.). Y.X. was supported by National Institutes of Health grant (K01 DK138281) and American Heart Association Training grant 827529. B.Z. was supported by Guangzhou Municipal Science and Technology Bureau SL2022A04J00542. Author contributions W.H., Y.X. and M.A.L. conceptualized the study, interpreted data and wrote the manuscript, which was revised and approved by all authors. B.Z., and B.C. performed MPRA and functional validation experiments. Y.X., Y.L., and J.Z. performed snATAC-seq. C.L. and N.H. performed bioinformatics analysis. X.C and F.Y. help with GWAS risks enrichment analysis. M.V., J.A.L., K.C., R.M., B.P. and D.J.R. performed PRS analysis. Data availability The accession number for the sequencing data reported in this paper are GSE281367, GSE281364, and GSE281160. Conflict of interests The authors declare the following competing interests: MAL is on a scientific advisory board of Pfizer and is co-founder and serves on the scientific advisory board of Flare Therapeutics. DJR is on the scientific advisory board for Alnylam, Novartis, and Verve. All other authors have no competing interests to declare. Methods Cell lines and culture Human hepatoma cell line HepG2 and Huh-7, human hepatic stellate cell line LX-2, and human embryonic kidney (HEK) 293T cell line were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum (FBS). Cells were maintained in a humidified incubator at 37 °C with 5% CO₂and supplemented with 100 U/mL penicillin and 100 µg/mL streptomycin. To establish an in vitro lipotoxic model, HepG2 or Huh-7 cells were treated with 0.15 mM palmitic acid (PA) and 0.3 mM oleic acid (OA). Both fatty acids were dissolved in 0.5% fatty acid-free bovine serum albumin (BSA). The treatment was administered for 24 h. To establish an in vitro fibrosis model, LX-2 cells were treated with 20 ng/mL transforming growth factor-beta 1 (TGFβ 1 ) for 48 h. Human iPSCs culture Human induced pluripotent stem cell (iPSC) lines were generated following the study protocol approved by The University of Pennsylvania Human Subjects Research Institutional Review Board, as previously reported 66 . iPSCs were cultured in mTeSR™ medium at 37 °C with 5% CO₂. Gene-edited stable monoclonal iPSC lines were produced as previously described 67 . In brief, human iPSCs were infected with Exci-sgRNA lentivirus co-expressing Cas9 and selected for 3 days using puromycin (0.5 µg/ml) and blasticidin (5 µg/ml). The surviving iPSCs were then subcloned at low seeding density and cultured for one week. Following colony isolation and expansion, Sanger sequencing was performed to identify stable DAV-deleted monoclonal human iPSC lines. Human studies Normal and MASLD human livers for snATAC-seq were provided by the Liver Tissue Cell Distribution System, Minneapolis, Minnesota, which was funded by National Institutes of Health contract no. HSN276201200017C. Additional human livers were obtained from SEKISUI XenoTech. All studies using human samples were exempted by the University of Pennsylvania Institutional Review Board. HepG2 CRISPRi/a cell lines construction The CRISPRi_HepG2 cell line was generated by targeted integration of a UCOE-CAG promoter-driven dCas9-KRAB/MeCP2-T2A-BSD expression cassette into the AAVS1 "safe harbor" site using CRISPR/Cas9. Donor plasmid (pUC19-AAVS1-dCas9-KRAB/MeCP2) and AAVS1 targeting plasmid (PX330-sgAAVS1_human) were co-transfected into HepG2 cells using Lipofectamine 3000 reagent. Blasticidin selection (30 µg/mL) commenced 72 h post-transfection and was maintained for at least 7 days. Expression of dCas9-KRAB/MeCP2 was evaluated by immunofluorescence. The CRISPRa_HepG2 cell line was established by infecting HepG2 cells with lentiviruses containing a reverse tetracycline-controlled trans-activator (rtTA) expression cassette co-expressing blasticidin S deaminase (BSD) and a 3G-Tet-regulated dCas9-VPR expression cassette. Blasticidin selection (30 µg/mL) was initiated 72 h post-infection and maintained for at least 7 days. Expression of dCas9-VPR was induced by treating cells with doxycycline (5 µg/mL) for 3 days and subsequently evaluated by immunofluorescence. Plasmids construction The donor plasmid (pUC19-AAVS1-dCas9-KRAB/MeCP2) was generated by cloning a UCOE-CAG-dCas9-KRAB/MeCP2-T2A-BSD expression cassette alongside with AAVS1 homology arm (HA) into pUC19 backbone. The HA-L, HA-R and UCOE element were amplified from human genome. The CAG promoter was amplified from SMARTvector. The dCas9-KRAB/MeCP2, T2A-BSD-WPRE and SV40-polyA fragments were amplified from the pLX-TRE-dCas9-KRAB/MeCP2-BSD vector. For PX330-sgAAVS1 plasmid, annealed oligo duplex (sgAAVS1_human) was prepared and cloned into BbsI cut pX330-U6-hSpCas9 backbone. The TetON-dCas9-VPR-BSD plasmid was generated by replacing the KRAB/MeCP2 expression sequence in the pLX-TRE-dCas9-KRAB/MeCP2-BSD vector with VPR expression sequence amplified from SP-dCas9-VPR plasmid. All the primers are shown in Supplemental Table S9. To produce lentiviruses, the lentivirus expression vector and the packaging plasmid mix (psPAX2 and pMD2.G) were co-transfected into HEK-293T cells. The lentivirus supernatant was harvested and stored in aliquots at -80°C until use. For pcDNA3.1- SLC22A3 overexpression (OE) plasmid construction. The human SLC22A3 CDS gene was obtained using HepG2 cDNA as a template by PCR amplification. The expression plasmid vector (pcDNA3.1+) was linearized by NheI and KpnI (NEB, Beijing, China) restriction enzyme digestion. The pcDNA3.1- SLC22A3 vector was constructed according to ClonExpress Ultra One Step Cloning Kit. Liver processing and nuclei isolation Isolation of nuclei for snATAC-seq was performed as previously described 17 . Briefly, frozen human liver tissue was homogenized using a Dounce homogenizer in lysis dilution buffer containing 10 mM Tris-HCl (pH 7.5), 10 mM NaCl, 3 mM MgCl₂, 1% BSA, and cOmplete™ Protease Inhibitor Cocktail. The homogenate was centrifuged, and the pellet was resuspended in lysis dilution buffer. To the nuclei suspension, lysis buffer (10 mM Tris-HCl, pH 7.5; 10 mM NaCl; 3 mM MgCl 2 ; 1% BSA; 0.1% Tween-20; 0.1% NP-40; 0.01% digitonin; cOmplete™ Protease Inhibitor Cocktail) was added. The nuclei were washed and resuspended in lysis dilution buffer. Nuclei were further purified using the Nuclei Isolation Kit with a 1.8 M sucrose cushion, following the manufacturer's instructions. After centrifugation at 30,000 g for 45 minutes at 4 °C, the nuclei pellet was collected and resuspended in PBS containing 1% BSA. The nuclei were filtered through a 40 μm cell strainer before loading onto the 10X Genomics Chromium Controller. snATAC-seq libraries construction and sequencing snATAC-seq libraries were prepared according to the manufacturer’s instructions of 10XGenomics (PN-1000176). Libraries were sequenced with Novogene (https://www.novogene.com/us-en/). 400 million paired-end reads were obtained for each sample using the sequencing parameters recommended in 10X Genomics guide. Cluster analysis of snATAC snATAC-seq data were preprocessed using the 10X Genomics pipeline (Cell Ranger ATAC v2.1.0) aligned to the hg38 genome assembly with default parameters. Samples from different patients were processed separately. High-quality individual cells were retained for downstream analysis (1,000 < unique barcode number 15%; blacklist ratio 3; nucleosome signal < 4). Quality control measurements and filtering were conducted using Signac (v. 1.12.9003) 68 . After quality control, all samples were merged into a single Seurat object. The integration function in Signac was then employed to remove batch effects. Unsupervised clustering was performed at a resolution of 0.4. Cell types were identified by integrating liver snRNA data with groups determined through unsupervised clustering. Motif activity was assessed using chromVAR (v. 1.24.0) 69 , and pseudotime scores were calculated with Monocle3 (v. 1.3.4). The pseudotime analysis was further confirmed using StaVia (v. 2.0). Gene Ontology (GO) analysis was conducted using GREAT (v. 4.0.4) 70 . Integrative analysis of snATAC-seq and snRNA-seq The liver snRNA data used for the integrative analysis was downloaded from GSE189600 17 . Cells were labeled according to the cell type information provided in the original publication. Subsequently, cells in the liver snATAC dataset were annotated by transferring cell labels from the liver snRNA dataset using tools from the Seurat (v. 5.0.2) package 71 . Utilizing the transferred labels and cell types identified through cluster-specific gene scores, correlation analyses were performed to assess the concordance between chromatin accessibility and gene expression profiles. Cell type-specific peaks and gene scores Cell type-specific peaks were identified by selecting significant peaks (adjusted p-value 0.58) that were not significant in more than three other cell types. Gene scores were calculated using the GeneActivity function in Signac (v1.12.9003) 68 , which quantifies the number of fragments per cell mapping to both the gene body and the promoter region (2 kb upstream). GWAS enrichment analysis by LDSC and SCAVENGE The LDSC package ( https://github.com/bulik/ldsc ) was used to perform LDSC analysis. GWAS summary statistics were downloaded from the GWAS Catalog: Triglycerides (GCST90092955), BMI (GCST90179150), and MASLD (GCST90054782). These GWAS summary statistics were converted to a format compatible with LDSC 72 using the provided script (munge_sumstats.py). For each cell type-specific peak, bed files were lifted over to hg19 from hg38 to create annotation files for each cell type. Coefficient and enrichment calculations were performed by LDSC following the cell-type-specific analysis tutorial ( https://github.com/bulik/ldsc/wiki/Cell-type-specific-analyses ). SCAVENGE analysis were perform as described 20 , GWAS summary statistics were downloaded from the GWAS Catalog for MASLD (GCST90091033), liver fat (GCST90016673) and cirrhosis (GCST90319878). Sentinel SNPs were identified using GCTA (v. 1.94.1) 73 with a significance threshold of p < 10 −8 . Fine-mapping analysis was conducted using coloc (v. 5.2.3) 74 with adj.P-value cutoff of 0.0001. The output from coloc was lifted over to hg19 from hg38 for downstream analysis. Information on counts, UMAP embeddings, and LSI embeddings was extracted from the results of the Signac analysis. Subsequently, SCAVENGE (v. 1.0.2) 20 utilized these data to compute the trait-relevant score (TRS). Joint analysis of single cell motif enrichment and GWAS enrichment Joint analysis of single-cell motif and GWAS enrichments was performed as previously described 10 , with minor modifications. Briefly, TF motif activity for each cell was quantified using chromVAR 69 , which infers TF activity by quantifying deviations in chromatin accessibility across sets of peaks sharing the same TF-binding motif. ChromVAR calculates a bias-corrected deviation score and corresponding Z-score by comparing the observed accessibility of motif-containing peaks to an expected accessibility derived from matched background peaks. Trait relevance scores (TRS) for individual cells were computed using SCAVENGE 20 . Spearman’s rank correlation coefficients were then calculated between motif activity and TRS across all cells to assess the association between transcription factor activity and GWAS trait enrichment. MPRA library design and construction MASLD-associated variants were obtained from the GWAS Catalog and relevant publications (Table S1). SNPs in Linkage disequilibrium (LD) with all collected significant SNPs were identified using the LDProxy function from LDlinkR, applying thresholds of minor allele frequency (MAF) > 5% and r² > 0.8 in European populations. Variants located within coding regions, as well as indels and multiallelic variants, were excluded from the analysis. For each remaining variant, a 126-base pair (bp) genomic sequence centered on the reference or alternate allele was extracted for oligonucleotide synthesis. In total, 5,369 variants were included for the MPRA library construction. The oligonucleotide library was designed and synthetized as 170 bp oligo pool, comprising a 126 bp genomic sequence (candidate regulatory sequence, CRS) flanked by 22 bp adapter sequences at both ends (5’-GAGTACTGTATGGGCGGGTACC [126 bp oligo] GCTAGCGACAAAAGTGTCAACT-3’). The MPRA plasmid library was constructed using the pGL4.24-miniP-GFP vector, where the luciferase expression sequence in the original pGL4.24 vector was replaced with a green fluorescent protein (GFP) expression sequence. A two-step PCR was employed to incorporate a minimal TATA promoter (miniP), unique 20 bp barcodes, and the necessary vector overhang sequences into the oligonucleotide library. The oligo pool PCR products were purified and subsequently cloned into the KpnI and NcoI digested pGL4.24-miniP-GFP vector using the ClonExpress Ultra One Step Cloning Kit as described above. The ligation reactions were purified and transformed into Lucigen Endura Electrocompetent Cells (Biosearch Technologies, 60242-2) via electroporation. The resulting plasmid library was purified using the ZymoPURE II™ Plasmid Maxiprep Kit. Finally, CRS-barcode fragments were amplified by PCR using the primers CRS/BC_F and CRS/BC_R for subsequent sequencing to establish CRS/barcode pairings. Massively parallel reporter assay The MPRA libraries were transfected into cells using Lipofectamine 3000 reagent. For HepG2 MPRA experiments, 2 × 10⁷ cells were used for one replicate. 16 h post-transfection, the medium was refreshed with complete culture medium with or without PAOA treatment and incubated for another 24 h. For LX-2 MPRA experiments, 1 × 10⁷ cells were used for one replicate. 6 h post-transfection, the medium was refreshed with complete culture medium with or without TGFβ treatment and incubated for an additional 36 h. A total of four biological replicates were performed for each experiment. Cells were harvested and performed DNA and RNA extraction using the Quick DNA/RNA Miniprep Kit following the manufacturer’s protocol. RNA samples are rigorous DNase treated using the RNase-free Recombinant DNase I Set. For each biological replicate, 240 µg of RNA was used to synthesize the first-strand cDNA library using SuperScript II in combination with a GFP mRNA-specific primer (MPRA_BCshort_R). To amplify random barcode from plasmid DNA or cDNA, and incorporate required P5/P7 flowcell sequence and sample index, the first-round PCR was performed using the primers MPRA_BClong_P5i5_F and MPRA_BClong_P7i7_F, followed by the second-round PCR using the primers P5 and P7. Each PCR reaction utilized 48 µg of DNA or total cDNA per biological replicate as the template. All PCR reactions were carried out using the NEBNext® High-Fidelity 2X PCR Master Mix. DNA and RNA barcode libraries were pooled in a 1:3 mole ratio. Library representation was assessed by next-generation sequencing on Illumina Novaseq platform. MPRA data analysis MPRA plasmid library Read 1 sequences were processed by trimming the fixed sequence preceding the inserted oligo sequences using cutadapt (https://doi.org/10.14806/ej.17.1.200) with the parameters --discard-untrimmed -m 80 -g GGCCTAACTGGCCGGTACCCTGAGTACTGTATGGGCGGGTACC, retaining only the subsequent sequences corresponding to the synthesized oligos and discarding reads shorter than 80 bp. Similarly, Read 2 sequences were trimmed to remove the fixed sequence preceding the inserted barcode using cutadapt with parameters --discard-untrimmed -l 20 -g TCACCATGGTGGCTTTACCAACAG, extracting the following 20 bp as barcode sequences. The first 107 bp of each synthesized oligo were extracted to create a reference genome, and Read 1 sequences were mapped to this reference using bwa mem 75 with parameters -L 100 -k 8 -O 5. Reads with a mapping score below 0.95 (mapping score = matched bases / total bases) were removed. Paired Read 2 sequences from the plasmid library corresponding to the retained Read 1 were used to construct oligo-barcode pairs, with duplicate pairs and barcodes mapping to multiple oligos further filtered to ensure unique pairings. For both cDNA and DNA libraries, fixed sequences were trimmed from the beginning of Read1 using cutadapt with parameters --discard-untrimmed -l 20 -g CTTGGCAATCCGGTACTTAGCGCATCTAAC. The reverse complement of the subsequent 20 bp sequences was treated as barcode sequences, and only barcodes present in the previously identified unique oligo-barcode pairs were retained for downstream analysis. Both DNA and cDNA barcodes in each oligo-barcode pair were quantified, and barcodes with a DNA barcode count of zero were excluded to ensure data quality. To assess oligo activity, we modified the MPRAprofiler protocol as described 76 , summarizing the total barcodes associated with each oligo for subsequent analyses. Potential enhancers were identified using DESeq2 77 to compare cDNA and DNA barcode counts for each oligo, with oligos exhibiting an adjusted p-value 0 (cDNA/DNA) classified as enhancers. Variants with either allele identified as active enhancers were further analyzed to determine differences in regulatory activity between the reference and alternative alleles. To detect differentially active variants, we first calculated the count ratio between the reference and alternative alleles for all variants. Subsequently, using the median ratio of four DNA replicates as a control, we normalized all RNA ratios and performed a Student’s t -test to calculate p-values. The Benjamini-Hochberg method was applied to adjust the p-values for multiple testing, and variants with an adjusted p-value < 0.01 were designated as significant differentially active variants. DAVs functional annotations DAVs were annotated to different genomic regions using ChIPseeker 78 . Additionally, published liver chromatin interaction data, eQTLs, and ChromHMM data were downloaded to elucidate the functional significance of DAVs. Overlaps between DAVs and chromatin loops were identified using the following datasets: (1) right lobe liver intact Hi-C data from ENCODE 79 (ID ENCFF961GKA); (2) HepG2 intact Hi-C data from ENCODE (ID ENCFF264RQT); (3) HepG2 POLR2A ChIA-PET data from ENCODE (ID ENCSR857MYZ); (4) capture-C data of promoters and H3K4me3/H3K27ac peaks in liver HepG2 cells 33 ; and (5) liver promoter capture Hi-C data from GSE86189 34 , including promoter-other interactions. Each loop anchor was extended by a fixed 2.5 kb on either side. DAVs with significant variant-gene pairs were annotated using liver eQTL data from GTEx v8 80 . Chromatin states of DAVs were predicted using the core 15-state model for human liver-related tissues and cells, as determined by ChromHMM 81 and downloaded from the ROADMAP Epigenetics Project 82 (IDs E066, E118). Eleven of these states were considered active and used for DAV annotation: 1_TssA, 2_TssAFlank, 3_TxFlnk, 4_Tx, 5_TxWk, 6_EnhG, 7_Enh, 8_ZNF/Rpts, 10_TssBiv, 11_BivFlnk, and 12_EnhBiv. The Bedtools 83 intersect command was utilized to calculate the overlap between the centers of DAVs and loop anchors, eQTLs, and ChromHMM-defined chromatin states. Enrichment of DAVs in cell type peaks To determine whether DAVs are preferentially associated with specific cell types, we assessed the overlap between the DAVs and cell type-specific peaks derived from scATAC-seq. This analysis involved comparing DAV overlaps with cell type-specific peaks against controls generated from 1,000 iterations of randomly selected genomic regions. Fisher’s exact test was employed to calculate p-values, determining the statistical significance of the observed overlaps. Additionally, we compared the number of overlaps between DAVs and all tested variants within specific cell type peaks to evaluate the enrichment tendency of DAVs relative to non-functional variants. We also analyzed the enrichment of all SNPs assayed by MPRA across promoter and enhancer elements in different cell types. Peaks located within promoter-proximal regions (-2 kb to +0.5 kb relative to transcription start sites) were classified as promoters, while all other peaks were designated as enhancers. Transcriptional factor binding analysis Transcription factor binding site (TFBS) disruptions by DAVs were assessed using SNP2TFBS and motifBreakR 84,85 . In the SNP2TFBS analysis, statistical significance was determined by comparing the proportion of DAVs predicted to disrupt a given TF motif to the proportion of background SNPs affecting the same motif. In contrast, the motifbreakR approach assessed significance by comparing the proportion of DAVs disrupting a TF motif to the proportion of all MPRA-tested SNPs with predicted effects on that TF. TFBS with an adjusted p-value < 0.01 were considered significantly affected by DAVs. Additionally, a more permissive approach was implemented to identify potentially disrupted TFBS by individual DAVs. This involved extracting the alternative and reference sequences centered on each variant, spanning ± 20 bp (totaling 41 bp). Motifs within these variant sequences were identified using MEME’s FIMO 86 against the HOCOMOCOv11_core_HUMAN database and HOMER TF database 87 , retaining only those motifs that encompassed the SNP location regardless of strand orientation. Subsequently, the position frequency matrix (PFM) values at the SNP position were compared between the reference and alternative alleles. Motifs with either the reference or alternative PFM value exceeding 0.3 and a ratio greater than 2 were designated as potentially interrupted motifs. For the exploration of differentially enriched TF motifs in DAVs and enhancers across various cell types and states, HOMER was utilized to identify motifs within condition-specific enhancers and DAVs, using other condition-specific enhancers or DAVs as background controls. Only motifs with a q-value < 0.1 were deemed enriched in DAVs and enhancers. For the significantly disrupted TFs by DAVs either in HepG2 or LX-2 cells, under both control and PAOA or TGFβ treatment, we constructed protein-protein interaction network by STRING 88 (https://cn.string-db.org/) with default parameter. Functional enrichment of these networks also displayed in WikiPathways with default parameter by STRING. Cell type-specific and stimuli-dependent enhancer and DAVs analysis Enhancers were identified separately across different cell types and states using DESeq2 77 by comparing cDNA and DNA barcode counts. Barcode counts for reference and alternative alleles were merged for each variant. Variants with an adjusted p-value 0 were classified as significant enhancers. To identify stimuli-dependent enhancers, only enhancers detected in either control or stimulation states were considered. Firstly, we compared the cDNA/DNA ratio of control versus stimulation in enhancers, normalizing by the median control ratio for each enhancer. P-values were corrected using the Benjamini–Hochberg method, and enhancers with a corrected p-value < 0.05 were designated as stimuli-responsive enhancers. Cell type-specific enhancers were identified using the same approach. DAVs identified across different cell types or states were overlapped; DAVs detected exclusively in one cell type were considered cell type-specific DAVs, and the same criterion was applied for different stimulation states. CROP-seq library design and experiments The CRISPR inhibition library (CROPi-DAVs) was designed and cloned as below. In brief, we designed sgRNAs to target 20 selected plausibly causal noncoding variants from our HepG2-MPRA screen (Supplementary Table 6). We extended 150 bp around center of each variant, then used FlashFry v1.15 89 and CRISPick 90 to detect potential sgRNAs. Finally, each variant was targeted by two sgRNAs from tops ranking by Hsu-Scott score. In addition, we included 6 non-targeting sgRNAs from the GeCKO v2 library as negative controls. All the sgRNAs sequences are shown in Supplemental Table S9. To create the CROPi-DAVs plasmid library, each sgRNAs oligos were annealed, and mixed together as an oligo duplexs pool. Then the oligo pool was cloned into BsmBI-v2 (NEB, R0739L) cut lenti-CROPseq-puro-v2 backbone. The purified ligation product was transformed into Competent-Cell Stbl3 (KTSM110L). Bacterial colonies were scraped in the following day, plasmids were isolated using the Endo-Free Plasmid Midi kit (CWBIO, CW2105S). For plasmid library QC, sgRNA sequences are amplified in a single 16-cycles PCR reaction with specially designed primers (CROP-P5i5-F and CROP-P7i7-R). Library representation was determined by Next-generation sequencing (Illumina). The CROPi-DAVs plasmid library was packaged into lentivirus as below. CROPi-DAVs plasmid library was co-transfected into HEK-293T cells with the packaging plasmid mix (psPAX2 and pMD2.G). Lentivirus was collected at 48h post-transfection, following by estimating multiplicity of infection (MOI) by counting cell survival ratio upon puromycin selection (6 μg/ml) in the HepG2 cell line. For library infection, CRISPRi_HepG2 cells seeded at 30% confluency in 10-cm dish were infected with CROPi-DAVs lentivirus library at a low MOI level (~0.1) to minimize double infection. Puromycin (6 μg/ml) screening was performed at 48h post-infection and following by maintaining for another 7 days. After puromycin selection and expansion, cells were captured by the 10X Chromium Controller using Chromium Next GEM Single Cell 3ʹ Reagent Kits v3.1 (10X Genomics, PN-1000268). Sample prep was performed according to protocol, holding 20~30 ng full-length cDNA per lane for sgRNA-enrichment PCR. Multiple PCR reactions were performed for sgRNA enrichment. In brief, in the first PCR reaction, full-length cDNA mix was used as template and 1st_erPCR_F and 1st_erPCR_R were used as primers. Then the first PCR product was cleaned up and amplified using the primers (2nd_erPCR_F and 2nd_erPCR_R). The final PCR was prepared in which 3rd_erPCR_F and 2nd_erPCR_R were used as primers. All PCR reactions were carried out for 10 cycles using 2x KAPA HiFi HotStart ReadyMix (KAPA Biosystems, KK2602) with annealing temperature at 62°C and 15 s extension per cycle. The whole-transcriptome scRNA-Seq libraries and the sgRNA-enrichment libraries were separately sequenced using novaXplus platform. CROP-seq analysis Cell-barcode assignment in the barcode-enrichment library was performed as previously described 91 . In brief, to map sgRNAs alongside other genomic transcripts in individual cells, we constructed a custom reference genome by adding pseudo-genes representing sgRNA-containing 500 bp transcripts to the human genome assembly (Ensembl GRCh38 release). Each pseudo-gene included one sgRNA sequence, with 250 bp upstream of the sgRNA containing a U6 promoter and 230 bp downstream containing LTR sequences. Both the genome sequence file and transcription GTF file were modified synchronously to create the final reference genome for the Cell Ranger pipeline. The barcode-enrichment library was sequenced with a 150 bp paired-end (150PE) strategy, where sgRNA-containing sequences were located in Read 2 (R2) and cell barcodes along with UMIs were in Read 1 (R1). To meet the file format requirements of the 10X Genomics Cell Ranger pipeline, the first 98 bp of R2 and the first 28 bp of R1 were retained using cutadapt (https://doi.org/10.14806/ej.17.1.200). Then, sequences were mapped to the custom reference genome using cellranger (10x Genomics, v.7.1.0). Subsequently, we utilized the get_barcodes.py script 91 to assign cell barcodes to sgRNAs. Initially, a whitelist of sgRNAs used in our library was created to comply with the requirements of get_barcodes.py. The script was then executed with the parameters --all_reads --search_seq CTTGTGGAAAGGACGAAACACCG, where the search sequence corresponds to the fixed sequence preceding the sgRNA. sgRNA targets were assigned to each cell based on dominance, requiring that a sgRNA be three times more abundant than the sum of all other sgRNAs within a cell; only unique assignments were retained. For the six lanes of single-cell transcription data from CROP-seq, each lane was individually mapped to the custom reference genome using cellranger (10x Genomics, v.7.1.0), and the resulting data were aggregated into a single dataset. The aggregated single-cell RNA-seq (scRNA-seq) data were processed using Seurat (v.4.2.0) 90 , with doublets removed by DoubletFinder (v.2.0.4) 92 . High-quality cells were retained following subsetting with criteria of nFeature_RNA > 200, nFeature_RNA < 10,000, and percent.mt < 10. The SCT normalization method was applied, and differentially expressed genes (DEGs) between sgRNA-targeting cells and other cells were identified using the FindMarkers function with the t-test method, utilizing the SCT assay and scale.data slot. Parameters were set to logfc.threshold = 0, min.cells.feature = 1, min.cells.group = 1, and min.pct = 0 to maximize the detection of potential DEGs. Target genes nomination analysis To explore the numerous potential target genes of DAVs, we employed four distinct methods that consider linear distance, chromatin interactions, co-accessibility, and gene expression effects. Firstly, proximal genes were identified by selecting genes whose transcription start sites (TSS) were located within 20 kb of DAVs. Secondly, utilizing the previously described chromatin interaction data, we identified loop target genes by extending each loop anchor to 5 kb. Genes with promoters (TSS ±500 bp) located within one anchor and DAVs situated within the other anchor of the same loop were classified as loop target genes. Specifically, for liver promoter capture Hi-C and H3K4me3/H3K27ac Capture-C data, which provided gene information within one anchor, DAVs were required to reside within the other anchor to be considered loop targets. Thirdly, we leveraged the run_cicero and generate_ccans functions from the Signac package (v.1.8.0) 68 to identify co-accessible sites from scATAC-seq data. Co-accessible sites with a score greater than 0.3 were retained, and each anchor was extended to 2 kb from the center. Genes with TSS located within one anchor and DAVs within the opposing anchor of the same co-accessible site were designated as DAVs’ co-access genes. Finally, genes associated with DAVs through significant variant-gene pairs from liver eQTL data were identified as DAV eQTL genes. These four approaches collectively considered linear proximity, chromatin interactions, co-accessibility of regulatory regions, and gene expression influences to comprehensively predict target genes regulated by DAVs. Inferring the gene regulatory network (GRN) The gene regulatory network (GRN) was implemented utilizing Pando (v1.0.0) 24 . Liver snRNA-seq 17 and snATAC-seq datasets were first integrated using Signac to enable joint modeling of gene expression and chromatin accessibility. GRNs were inferred separately for MASLD and normal hepatocytes to characterize disease-associated regulatory rewiring. The infer_grn() function in Pando was applied with default parameters, using all accessible chromatin regions and expressed genes as input features. To identify transcription factor (TF)-based regulatory modules, the find_modules() function was employed to scan all accessible regions for TF-binding motifs based on the curated motif database embedded in Pando. Peaks containing TF motifs were linked to nearby genes using the “Signac” method, and the regulatory model was fitted using the “glm” method. TF modules were subsequently constructed with default settings, and only the highest-confidence interactions for each TF were retained. Visualization of the reconstructed GRN was performed using the NetworkGraph() function, and networks were rendered with ggraph (v2.2.2.9000). In the network visualization, node color represents the log₂ fold change in gene expression between MASLD and normal hepatocytes, node size reflects PageRank centrality (a measure of network importance), and edge color encodes the statistical significance of each regulatory connection within its assigned module. To delineate regulatory relationships associated with disease-associated variant (DAV) target genes, a comprehensive GRN was first reconstructed using all accessible regions and genes across cell types, followed by extraction of a DAV-focused subnetwork containing only regulatory connections linked to predefined DAV target genes. We further profiled TF expression dynamics and motif activity within hepatocyte GRNs. TF expression levels in MASLD and normal hepatocytes were obtained using the aggregate_matrix() function in Pando. Motif activity was computed by first annotating motifs with the AddMotifs() function from Signac, followed by activity quantification using the RunChromVAR() function. Average motif activity for each TF was obtained using the wilcoxauc() function and compared between MASLD and normal states to identify differentially active TFs within the reconstructed GRN. Luciferase reporter assay CRS oligos consisting of reference/alternative (ref/alt) pairs were selected based on our MPRA screen. Each 350 bp CRS sequence was synthesized as a gBlock and subsequently cloned into the pGL4.24 firefly luciferase reporter vector (Promega, E8421). Standard luciferase reporter assays were performed in HepG2, LX-2, Huh7 or HEK 293T cells. 36 h post-transfection, luciferase activity was quantified using the Dual-Luciferase® Reporter Assay System according to the manufacturer’s instructions. Luminescence signals were measured in a white 384-well plate using a BioTek Synergy Neo2 plate reader. The activity of firefly luciferase was normalized to Renilla luciferase activity to assess the functional activity of the tested regulatory elements. CRIPSR deletion, inhibition and activation For CRIPSR deletion sgRNA (Exci-sgRNA) plasmid construction, sgRNAs targeting the variants of interest were designed using CRISPOR. Exci-sgRNA oligos were annealed and ligated into the lentiCRISPR-v2-Puro vector or lentiCRISPR-v2-Blast vector, respectively. For CRIPSR inhibition (CRISPRi) and activation (CRISPRa) sgRNA plasmid construction, sgRNAs targeting the variants of interest were designed using CRISPick, the sgRNAs oligo inserts were annealed and cloned into the lenti-CROPseq-puro-v2 vector as described above. The lentivirus particles were produced, harvested and stored in aliquots at -80°C until use. For CRIPSR deletion, HepG2 cells were infected with Exci-sgRNA lentivirus and selected for 7 days using puromycin (6 µg/ml) and blasticidin (30 µg/ml). The deletion effect of each targeted variant was estimated by PCR amplification using primers targeting outside of the deletion sequence (Supplemental Table S9). The deleted small genomic regions were detected using agarose gel electrophoresis. For CRISPR inhibition, the CRISPRi_HepG2 cells were infected with CRISPRi-sgRNA lentivirus followed by puromycin selection (6 μg/ml) for at least 7 days. For CRISPR activation, the CRISPRa_HepG2 cells were infected with CRISPRa-sgRNA lentivirus followed by puromycin selection (6 μg/ml) and doxycycline treatment (5 µg/mL) for at least 7 days. All the sgRNAs sequences and PCR amplification primers are shown in Supplemental Table S9. Gene editing in HepG2 PE5b system was used for SNP editing in HepG2 following the protocol as previously described 93 . The epegRNA and nickRNA was designed to recognize the surrounding region of the SNP using PRIDICT2.0 ( https://www.pridict.it/ ), then the epegRNA dsODN for rs474513(G>A) and nickRNA dsODN for rs474513 were synthetized and cloned into tevopreq1 epegRNA vector and nick-sgRNA vector respectively. The rs474513(G>A) epegRNA plasmid, rs474513 nick-sgRNA plasmid, pCMV-PEmax-P2A-GFP and pEF1a-MLH1dn were co-transfected into HepG2 using Lipofectamine 3000 reagent. 48 h post-transfection, the GFP-positive cells were sorted by FACS and the subclones were identified by Sanger sequencing. RT-qPCR Cells were collected and RNA was isolated with RNeasy Mini Kit according to the manufacturer’s instructions. For mRNA detection, RNA samples were reverse transcribed using HiScript® III RT SuperMix for qPCR (+gDNA wiper), and qRT-PCR was performed with ChamQ Universal SYBR qPCR Master Mix using Applied Biosystems QuantStudio 5 (Thermo Fisher Scientific, USA). Sequences of the qPCR primers are shown in Supplemental Table S9. Relative quantification analysis was performed using the ΔΔCt method, with HPRT1 as an endogenous reference. Relative gene expression is presented as the ratio of the target gene to reference control. Hepatic-like cells differentiation The monoclonal human iPSC lines were differentiated to hepatic-like cells as previously described 94 . In brief, hiPSCs were differentiated into Definitive Endoderm (DE) using STEMdiff Definitive Endoderm Kit for 4 days following the manufacturer’s instructions. For 4-9 days, cells were cultured in Advanced DMEM/F12 with 1% B27 (Cat: #A1895601), 1% KSR, 1% Glutamax, 1% MEM-NEAA, 0.5 µM A83-01, 0.25 µM Na-butyrate and 0.5% DMSO. Cells were maintained at 37°C in 5% CO 2 with 95% air and the medium was replaced daily. For 9-16 days, cells were cultured in Advanced DMEM/F12 with 1% B27 (Cat: #A1895601), 1% KSR, 1% Glutamax, 1% MEM-NEAA, 0.5 µM A83-01, 15 µM FH1, 15 µM FPH1, 0.1 µM Dexamethasone and 10 µM Hydrocortisone. Cells were maintained at 37°C in 5% CO 2 with 95% air and the medium was replaced daily. Cells were collected at Day0, Day4, Day9, Day14 and Day16, and lysed for RNA extraction and RT-qPCR detection. Hepatic organoid (HLO) differentiation The hepatic organoid (HLO) differentiation was performed as previously described 95 . In brief, hiPSCs were differentiated into Definitive Endoderm (DE) using STEMdiff Definitive Endoderm Kit for 4 days following the manufacturer’s instructions. For 4–6 days, cells were cultured in Advanced DMEM/F12 with 2% B27 (Cat#17504044) and 1% N2 containing 500 ng/mL fibroblast growth factor 4 (FGF4) and 3 µM CHIR99021. Cells were maintained at 37°C in 5% CO 2 with 95% air and the medium was replaced daily. The FG cells were detached by Accutase and suspended with Matrigel matrix on ice. The mixture of cells and Matrigel was embedded in 30 µL drops on dishes in Advanced DMEM/F12 with 2% B27, 1% N2, 10 mM HEPES, 1% Glutamax, 1% Pen/Strep, 5 ng/mL fibroblast growth factor 2 (FGF2) , 10 ng/mL vascular endothelial growth factor (VEGF), 20 ng/mL epidermal growth factor (EGF), 3 µM CHIR99021, 0.5 µMA83-01, and 50 mg/mL ascorbic acid, and incubated in the CO 2 incubator for 4 days with medium changed every 2 days. The medium was then switched to Advanced DMEM/F12 with 2% B27 (Cat#17504044), 1% N2, 10 mM HEPES, 1% Pen/Strep, and 2 µM retinoic acid (RA), and incubated for further 4 days with medium changed every 2 days. The final media switch was to the hepatocyte culture medium (HCM) and the cells were incubated for another 6 days, changing the medium every 2 days. Fatty liver organoid HLOs model Modeling of steatohepatitis in HLOs was performed as previously reported 95 . HLOs was isolated from Matrigel and washed with cold PBS in 4°C for three times, then cultured with HCM media containing 5 µg/mL insulin and 300 µM oleic acid and 150 µM palmitic acid on ultra-low attachment 6 multi-well plates (Corning) to induce steatohepatitis. HLOs were collected at day 4 for immunostaining and RNA isolation. Accumulation of lipid in HLOs was measured using BODIPY 493/503. 3D Liver Spheroids The HepG2/LX-2 spheroids were generated following the protocol described previously 49 . HepG2-Exci-sgRNA stable cell lines and LX-2 cell lines were mixed in a 24:1 ratio into 96-well round-bottomed ultra-low attachment plates (Corning) at 2000 cells/well in 100 μL minimum essential medium supplemented with 10% fetal bovine serum (FBS). After 72 h, they were treated with 300 µM oleic acid and 150 µM palmitic acid for another 48h. Immunostaining For HepG2 cell lines and HepG2/LX-2 3D spheroids immunostaining, cells were washed twice with PBS and fixed with 4% Paraformaldehyde (PFA) for 15min at room temperature. After this, samples were stained with 5µM BODIPY (493/503 nm) and 1 µg/ml DAPI for 15 min at 37°C. For LX-2 cell lines immunostaining, cells were washed twice with PBS and fixed with 4% Paraformaldehyde (PFA) for 15 min at room temperature. After this, samples were then permeated with 0.1% Triton X-100/PBS for 5 min and blocked with 10% BSA/PBS to avoid nonspecific staining for 1 h. Samples were reacted with primary antibodies diluted in 5% BSA at 4°C overnight, following by incubation with fluorophore-labeled secondary antibodies (4 μg/mL) diluted in 5% BSA/PBS at 37°C for 30 min, followed by staining DAPI (1 μg/mL) diluted in PBS at room temperature for 15 min. For HLOs immunostaining, HLOs were collected and washed twice with PBS and fixed with 4% PFA for 2 h at room temperature. Then, HLOs were permeabilized with 0.5% Triton X-100/PBS for 20 min at room temperature and blocked with PBS containing 0.5% Triton X-100 and 1% bovine serum albumin (BSA) for another 2 h at room temperature. HLOs were then stained with specific antibodies or 5 µM BODIPY (493/503 nm) and 1 µg/ml DAPI. After with washing with three times of DPBS containing 0.5% Triton X-100 and 0.2% BSA, HLOs were ready for imaging. All samples were imaged using confocal microscope (Nikon A1). FACS HepG2 cell lines were detached by trypsin and wash with PBS for 3 times, following by 4% PFA fixation for 15 min at room temperature. After this, samples were stained with 5 µM BODIPY (493/503 nm) and 1µg/ml DAPI for 1.5 h at 37°C. Cells were washed in PBS and resuspended in PBS and transferred into Test Tube with Cell Strainer Snap Cap. Analysis of cells was performed using Agilent Novocyte Advanteon. Polygenic risk score analysis The PRS analysis was conducted using available data from the VA’s Million Veteran Program (MVP) 60 and initial MASLD-related analyses within VA Merit-funded MVP Cardiometabolic Project (MVP003/028) 4 . MVP is a large-scale biobank, supported by the Veterans Health Administration Office of Research and Development in the United States. The study protocols of MVP biobank and MVP Cardiometabolic Project were approved by Veterans Affairs (VA) Central Institutional Review Board (CIRB). The design, baseline demographics, and quality control procedures for MVP biobank have been previously described 60 . The MASLD-related phenotype and genetic analyses in MVP Cardiometabolic Projects formed the basis for the current PRS analyses 4 . PRS was constructed to test the MASLD predictive power of SNPs prioritized by MPRA. After removing the SNPs in linkage disequilibrium (r2 > 0.05), the 114 independent SNPs out of the 365 SNPs prioritized by MPRA assays were used to build PRS. A matched set of 114 SNPs, selected from meta-analyzed results of 10 different randomized draws of MPRA total SNPs, excluding 365 DAVs, served as the negative control. A set of 9 missense SNPs (PRS9; Table S8), which are the well-established MASLD causal SNPs 4 , served as the positive control. Four different MASLD phenotype proxies were used to identify MASLD: MASLD-ALT, MASLD-ICD, ALT and FIB4 max progression. (1) For MASLD-ALT, MASLD cases were defined as participants with elevated ALT (> 40 U/L for men or >30 U/L for women during at least two time points at least 6 months apart within a 2-year window at any point prior to enrollment and exclusion of other causes of liver disease, chronic liver diseases or systemic conditions and/or alcohol use disorders. (2) For MASLD-ICD, MASLD cases were identified by ICD9 codes (456.2, 456.21, 571.5, 572.2 and 572.3) and ICD-10 codes (K72.9, K72.91, K74.0, K74.02, K74.1, K74.2, K74.6 and K74.69). (3) The MASLD phenotype proxy, ALT, was a continuous variable representing ALT levels. (4) Fibrosis-4 (FIB4) max progression was a surrogate marker of liver fibrosis progression by estimating FIB4 trajectory over time 96 . FIB4 was calculated using the formula ([age × AST]/[platelet × sqrt (ALT)]) as previously described. AST, ALT, and platelet count (in K/mm3) of all outpatient were collected from VA laboratories from January 2002 to November 2021. To construct the PRS, logistic regression was performed between SNPs and MASLD proxies MASLD-ALT, MASLD-ICD, or FIB-4 max progression, while linear regression was used between SNPs and ALT levels. Variant weights were derived from effect sizes reported in the GWAS by Miao et al 7 . The regression analysis of PRS with MASLD phenotypes was performed using a linear regression model adjusted for age, sex, and principal components of genetic ancestry. The regression coefficients from these analyses represented log-odds change in the outcome for each unit of increase in the standardized PRS. To compare the predictive power in deciles, the individuals were divided into 10 deciles based on their PRS scores and assessed the odds ratio (OR) of having MASLD (MASLD-ALT) in each of the 10 deciles when compared to the lowest decile. Comparison of DAV-derived PRS with published PRS models The polygenic risk score (PRS) constructed from disease-associated variants (DAVs) was benchmarked against the best-performing published MASLD PRS (PGS002732) derived from the MVP cohort, as well as PRS9 and PGS000655 4,27 . To minimize potential overfitting, all PRS models were evaluated in the All of Us Research Program (AoU)as an independent validation cohort. AoU is a large, US-based biobank developed to improve and enable the inclusion of diverse, underrepresented populations in large-scale genetic and epidemiological research studies 97 . We accessed short read whole genome sequencing (srWGS) data, survey data, and electronic health record data from participants included in the v8 release of AoU. Age within AoU was calculated as participants’ age as of their consent date to AoU. Sex was determined using self-reported sex-at-birth data provided by AoU. We used MASLD-ALT as a proxy for MASLD, as defined above. ALT labs in units per liter were extracted using LOINC code 1742-6. Measurements less than 0 were removed, then outlier measurements were removed if they were 3 standard deviations away from the population average in log space. DAV PRS for MASLD were calculated using the 114 LD-independent DAVs and marginal effect sizes from the Vujkovic et al . GWAS 4 . Published PRS were downloaded from the PGSCatalog 98 . All PRS were calculated using PLINK v1 with the --score sum argument. We residualized global ancestry from the PRS as previously described 99 , then standardized the scores. Association analyses were adjusted for age, sex, the first 10 PCs, and sequencing site. Statistical differences in effect size were estimated using a two-sided Z-test. P-values were adjusted for multiple testing per analysis using the Benjamini-Hochberg procedure. Quantification and statistical analysis Data are presented as mean ± SEM. Statistical significance was determined by unpaired two-tailed student t-test, Fisher’s exact test or two-way anova analysis using GraphPad Prism software or R software; a p-value of < 0.05 or FDR<0.01 was considered significant. Specific statistical tests performed are listed in the respective figure legends or sections of the Methods. References Matchett, K.P., Paris, J., Teichmann, S.A. & Henderson, N.C. Spatial genomics: mapping human steatotic liver disease. Nat Rev Gastroenterol Hepatol 21, 646–660 (2024). Eslam, M. & George, J. Genetic contributions to NAFLD: leveraging shared genetics to uncover systems biology. Nat Rev Gastroenterol Hepatol 17, 40–52 (2020). Chen, Y. et al. Genome-wide association meta-analysis identifies 17 loci associated with nonalcoholic fatty liver disease. Nat Genet 55, 1640–1650 (2023). Vujkovic, M. et al. A multiancestry genome-wide association study of unexplained chronic ALT elevation as a proxy for nonalcoholic fatty liver disease with histological and radiological validation. Nat Genet 54, 761–771 (2022). Romeo, S. et al. Genetic variation in PNPLA3 confers susceptibility to nonalcoholic fatty liver disease. Nat Genet 40, 1461–5 (2008). Anstee, Q.M. et al. Genome-wide association study of non-alcoholic fatty liver and steatohepatitis in a histologically characterised cohort(☆). J Hepatol 73, 505–515 (2020). Miao, Z. et al. Identification of 90 NAFLD GWAS loci and establishment of NAFLD PRS and causal role of NAFLD in coronary artery disease. HGG Adv 3, 100056 (2022). Zeng, B. et al. Genetic regulation of cell type-specific chromatin accessibility shapes brain disease etiology. Science 384, eadh4265 (2024). Cooper, Y.A. et al. Functional regulatory variants implicate distinct transcriptional networks in dementia. Science 377, eabi8654 (2022). Chiou, J. et al. Single-cell chromatin accessibility identifies pancreatic islet cell type- and state-specific regulatory programs of diabetes risk. Nat Genet 53, 455–466 (2021). Gaulton, K.J., Preissl, S. & Ren, B. Interpreting non-coding disease-associated human variants using single-cell epigenomics. Nat Rev Genet 24, 516–534 (2023). Nott, A. et al. Brain cell type-specific enhancer-promoter interactome maps and disease-risk association. Science 366, 1134–1139 (2019). Gordon, M.G. et al. lentiMPRA and MPRAflow for high-throughput functional characterization of gene regulatory elements. Nat Protoc 15, 2387–2412 (2020). Deng, C. et al. Massively parallel characterization of regulatory elements in the developing human cortex. Science 384, eadh0559 (2024). Rummel, C.K. et al. Massively parallel functional dissection of schizophrenia-associated noncoding genetic variants. Cell 186, 5165–5182 e33 (2023). Seidman, J.S. et al. Niche-Specific Reprogramming of Epigenetic Landscapes Drives Myeloid Cell Diversity in Nonalcoholic Steatohepatitis. Immunity 52, 1057–1074 e7 (2020). Xiao, Y. et al. Hepatocytes demarcated by EphB2 contribute to the progression of nonalcoholic steatohepatitis. Sci Transl Med 15, eadc9653 (2023). Kim, H.Y. et al. Multi-modal analysis of human hepatic stellate cells identifies novel therapeutic targets for metabolic dysfunction-associated steatotic liver disease. J Hepatol 82, 882–897 (2025). Govaere, O. et al. Transcriptomic profiling across the nonalcoholic fatty liver disease spectrum reveals gene signatures for steatohepatitis and fibrosis. Sci Transl Med 12(2020). Yu, F. et al. Variant to function mapping at single-cell resolution through network propagation. Nat Biotechnol 40, 1644–1653 (2022). Wiering, L., Subramanian, P. & Hammerich, L. Hepatic Stellate Cells: Dictating Outcome in Nonalcoholic Fatty Liver Disease. Cell Mol Gastroenterol Hepatol 15, 1277–1292 (2023). Wang, P., Kang, Q., Wu, W.S. & Rui, L. Hepatic Snai1 and Snai2 promote liver regeneration and suppress liver fibrosis in mice. Cell Rep 43, 113875 (2024). Yang, Z.F., Drumea, K., Mott, S., Wang, J. & Rosmarin, A.G. GABP transcription factor (nuclear respiratory factor 2) is required for mitochondrial biogenesis. Mol Cell Biol 34, 3194–201 (2014). Fleck, J.S. et al. Inferring and perturbing cell fate regulomes in human brain organoids. Nature 621, 365–372 (2023). Fairfield, C.J. et al. Genome-Wide Association Study of NAFLD Using Electronic Health Records. Hepatol Commun 6, 297–308 (2022). Ghodsian, N. et al. Electronic health record-based genome-wide meta-analysis provides insights on the genetic architecture of non-alcoholic fatty liver disease. Cell Rep Med 2, 100437 (2021). Namjou, B. et al. GWAS and enrichment analyses of non-alcoholic fatty liver disease identify new trait-associated genes and pathways across eMERGE Network. BMC Med 17, 135 (2019). Gao, C. et al. Genome-wide association analysis of serum alanine and aspartate aminotransferase, and the modifying effects of BMI in 388k European individuals. Genet Epidemiol 45, 664–681 (2021). Emdin, C.A. et al. Association of Genetic Variation With Cirrhosis: A Multi-Trait Genome-Wide Association and Gene-Environment Interaction Study. Gastroenterology 160, 1620–1633 e13 (2021). Emdin, C.A. et al. A missense variant in Mitochondrial Amidoxime Reducing Component 1 gene and protection against liver disease. PLoS Genet 16, e1008629 (2020). Gallagher, M.D. & Chen-Plotkin, A.S. The Post-GWAS Era: From Association to Function. Am J Hum Genet 102, 717–730 (2018). Harris, L. et al. Genome-wide association testing beyond SNPs. Nat Rev Genet 26, 156–170 (2025). Caliskan, M. et al. Genetic and Epigenetic Fine Mapping of Complex Trait Associated Loci in the Human Liver. Am J Hum Genet 105, 89–107 (2019). Jung, I. et al. A compendium of promoter-centered long-range chromatin interactions in the human genome. Nat Genet 51, 1442–1449 (2019). Musso, G., Cassader, M., Bo, S., De Michieli, F. & Gambino, R. Sterol regulatory element-binding factor 2 (SREBF-2) predicts 7-year NAFLD incidence and severity of liver disease and lipoprotein and glucose dysmetabolism. Diabetes 62, 1109–20 (2013). Kang, B., Kang, B., Roh, T.Y., Seong, R.H. & Kim, W. The Chromatin Accessibility Landscape of Nonalcoholic Fatty Liver Disease Progression. Mol Cells 45, 343–352 (2022). Liu, D. et al. Ets-1 deficiency alleviates nonalcoholic steatohepatitis via weakening TGF-beta1 signaling-mediated hepatocyte apoptosis. Cell Death Dis 10, 458 (2019). Mao, Z. et al. ETV5 Regulates Hepatic Fatty Acid Metabolism Through PPAR Signaling Pathway. Diabetes 70, 214–226 (2021). Isaac, R. et al. TM7SF3 controls TEAD1 splicing to prevent MASH-induced liver fibrosis. Cell Metab 36, 1030–1043 e7 (2024). Buck, M. & Chojkier, M. C/EBPbeta associates with caspase 8 complex proteins and modulates apoptosis in hepatic stellate cells. J Clin Gastroenterol 41 Suppl 3, S295–9 (2007). Hatting, M. et al. Hepatocyte caspase-8 is an essential modulator of steatohepatitis in rodents. Hepatology 57, 2189 – 201 (2013). Wang, L. et al. The multiple roles of interferon regulatory factor family in health and disease. Signal Transduct Target Ther 9, 282 (2024). Abul-Husn, N.S. et al. A Protein-Truncating HSD17B13 Variant and Protection from Chronic Liver Disease. N Engl J Med 378, 1096–1106 (2018). Cherubini, A. et al. Interaction between estrogen receptor-alpha and PNPLA3 p.I148M variant drives fatty liver disease susceptibility in women. Nat Med 29, 2643–2655 (2023). Pearson-Gallion, B. et al. Abstract 321: Altered Epha2 Signaling In Non-alcoholic Fatty Liver Disease Progression. 43, A321-A321 (2023). Datlinger, P. et al. Pooled CRISPR screening with single-cell transcriptome readout. Nat Methods 14, 297–301 (2017). Yang, Y., Konrad, R.J., Ploug, M. & Young, S.G. APOA5 deficiency causes hypertriglyceridemia by reducing amounts of lipoprotein lipase in capillaries. J Lipid Res 65, 100578 (2024). Liu, J., Afroza, H., Rader, D.J. & Jin, W. Angiopoietin-like protein 3 inhibits lipoprotein lipase activity through enhancing its cleavage by proprotein convertases. J Biol Chem 285, 27561–70 (2010). Pennisi, G. et al. ANGPTL3 Downregulation Increases Intracellular Lipids by Reducing Energy Utilization. Arterioscler Thromb Vasc Biol 44, 1086–1097 (2024). Bergmark, B.A. et al. Effect of Vupanorsen on Non-High-Density Lipoprotein Cholesterol Levels in Statin-Treated Patients With Elevated Cholesterol: TRANSLATE-TIMI 70. Circulation 145, 1377–1386 (2022). Nevzorova, Y.A. et al. Enhanced expression of c-myc in hepatocytes promotes initiation and progression of alcoholic liver disease. J Hepatol 64, 628–40 (2016). Ress, C. et al. The role of apolipoprotein A5 in non-alcoholic fatty liver disease. Gut 60, 985 – 91 (2011). Ottensmann, L. et al. Genome-wide association analysis of plasma lipidome identifies 495 genetic associations. Nat Commun 14, 6934 (2023). Teratani, T. et al. Lipoprotein Lipase Up-regulation in Hepatic Stellate Cells Exacerbates Liver Fibrosis in Nonalcoholic Steatohepatitis in Mice. Hepatol Commun 3, 1098–1112 (2019). Pan, Z. et al. Inhibition of MERTK reduces organ fibrosis in mouse models of fibrotic disease. Sci Transl Med 16, eadj0133 (2024). Mahajan, A. et al. Fine-mapping type 2 diabetes loci to single-variant resolution using high-density imputation and islet-specific epigenome maps. Nat Genet 50, 1505–1513 (2018). Bertran, L. et al. The Potential Protective Role of RUNX1 in Nonalcoholic Fatty Liver Disease. Int J Mol Sci 22(2021). Ouchi, R. et al. Modeling Steatohepatitis in Humans with Pluripotent Stem Cell-Derived Organoids. Cell Metab 30, 374–384 e6 (2019). Kachuri, L. et al. Principles and methods for transferring polygenic risk scores across global populations. Nat Rev Genet 25, 8–25 (2024). Gaziano, J.M. et al. Million Veteran Program: A mega-biobank to study genetic influences on health and disease. J Clin Epidemiol 70, 214 – 23 (2016). Huang, X., Rymbekova, A., Dolgova, O., Lao, O. & Kuhlwilm, M. Harnessing deep learning for population genetic inference. Nat Rev Genet 25, 61–78 (2024). Hutchinson, A., Asimit, J. & Wallace, C. Fine-mapping genetic associations. Hum Mol Genet 29, R81-R88 (2020). Lee, D. et al. A method to predict the impact of regulatory variants from DNA sequence. Nat Genet 47, 955–61 (2015). Benaglio, P. et al. Mapping genetic effects on cell type-specific chromatin accessibility and annotating complex immune trait variants using single nucleus ATAC-seq in peripheral blood. PLoS Genet 19, e1010759 (2023). Hemani, G., Bowden, J. & Davey Smith, G. Evaluating the potential role of pleiotropy in Mendelian randomization studies. Hum Mol Genet 27, R195-R208 (2018). Pashos, E.E. et al. Large, Diverse Population Cohorts of hiPSCs and Derived Hepatocyte-like Cells Reveal Functional Genetic Variation at Blood Lipid-Associated Loci. Cell Stem Cell 20, 558–570 e10 (2017). Hu, W. et al. Patient Adipose Stem Cell-Derived Adipocytes Reveal Genetic Variation that Predicts Antidiabetic Drug Response. Cell Stem Cell 24, 299–308 e6 (2019). Stuart, T., Srivastava, A., Madad, S., Lareau, C.A. & Satija, R. Single-cell chromatin state analysis with Signac. Nat Methods 18, 1333–1341 (2021). Schep, A.N., Wu, B., Buenrostro, J.D. & Greenleaf, W.J. chromVAR: inferring transcription-factor-associated accessibility from single-cell epigenomic data. Nat Methods 14, 975–978 (2017). Tanigawa, Y., Dyer, E.S. & Bejerano, G. WhichTF is functionally important in your open chromatin data? PLoS Comput Biol 18, e1010378 (2022). Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587 e29 (2021). Bulik-Sullivan, B.K. et al. LD Score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat Genet 47, 291–5 (2015). Yang, J., Lee, S.H., Goddard, M.E. & Visscher, P.M. GCTA: a tool for genome-wide complex trait analysis. Am J Hum Genet 88, 76–82 (2011). Giambartolomei, C. et al. Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. PLoS Genet 10, e1004383 (2014). Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–60 (2009). Lu, X. et al. Global discovery of lupus genetic risk variant allelic enhancer activity. Nat Commun 12, 1611 (2021). Love, M.I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15, 550 (2014). Yu, G., Wang, L.G. & He, Q.Y. ChIPseeker: an R/Bioconductor package for ChIP peak annotation, comparison and visualization. Bioinformatics 31, 2382–3 (2015). Sloan, C.A. et al. ENCODE data at the ENCODE portal. Nucleic Acids Res 44, D726-32 (2016). Consortium, G.T. The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science 369, 1318–1330 (2020). Ernst, J. & Kellis, M. ChromHMM: automating chromatin-state discovery and characterization. Nat Methods 9, 215–6 (2012). Bernstein, B.E. et al. The NIH Roadmap Epigenomics Mapping Consortium. Nat Biotechnol 28, 1045–8 (2010). Quinlan, A.R. BEDTools: The Swiss-Army Tool for Genome Feature Analysis. Curr Protoc Bioinformatics 47, 11 12 1–34 (2014). Kumar, S., Ambrosini, G. & Bucher, P. SNP2TFBS - a database of regulatory SNPs affecting predicted transcription factor binding site affinity. Nucleic Acids Res 45, D139-D144 (2017). Coetzee, S.G., Coetzee, G.A. & Hazelett, D.J. motifbreakR: an R/Bioconductor package for predicting variant effects at transcription factor binding sites. Bioinformatics 31, 3847–9 (2015). Bailey, T.L., Johnson, J., Grant, C.E. & Noble, W.S. The MEME Suite. Nucleic Acids Res 43, W39-49 (2015). Heinz, S. et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol Cell 38, 576–89 (2010). Szklarczyk, D. et al. The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Res 51, D638-D646 (2023). McKenna, A. & Shendure, J. FlashFry: a fast and flexible tool for large-scale CRISPR target design. BMC Biol 16, 74 (2018). Doench, J.G. et al. Optimized sgRNA design to maximize activity and minimize off-target effects of CRISPR-Cas9. Nat Biotechnol 34, 184–191 (2016). Hill, A.J. et al. On the design of CRISPR-based single-cell molecular screens. Nat Methods 15, 271–274 (2018). McGinnis, C.S., Murrow, L.M. & Gartner, Z.J. DoubletFinder: Doublet Detection in Single-Cell RNA Sequencing Data Using Artificial Nearest Neighbors. Cell Syst 8, 329–337 e4 (2019). Doman, J.L., Sousa, A.A., Randolph, P.B., Chen, P.J. & Liu, D.R. Designing and executing prime editing experiments in mammalian cells. Nat Protoc 17, 2431–2468 (2022). Hu, W. et al. Individual-specific functional epigenomics reveals genetic determinants of adverse metabolic effects of glucocorticoids. Cell Metab 33, 1592–1609 e7 (2021). Kimura, M. et al. En masse organoid phenotyping informs metabolic-associated genetic susceptibility to NASH. Cell 185, 4216–4232 e16 (2022). Kaplan, D.E. et al. Clinical and genetic risk factors for progressive fibrosis in metabolic dysfunction-associated steatotic liver disease. Hepatol Commun 8(2024). All of Us Research Program Genomics, I. Genomic data in the All of Us Research Program. Nature 627, 340–346 (2024). Lambert, S.A. et al. The Polygenic Score Catalog as an open database for reproducibility and systematic evaluation. Nat Genet 53, 420–425 (2021). Khera, A.V. et al. Whole-Genome Sequencing to Characterize Monogenic and Polygenic Contributions in Patients Hospitalized With Early-Onset Myocardial Infarction. Circulation 139, 1593–1602 (2019). Additional Declarations Yes there is potential Competing Interest. MAL is on a scientific advisory board of Pfizer and is co-founder and serves on the scientific advisory board of Flare Therapeutics. DJR is on the scientific advisory board for Alnylam, Novartis, and Verve. All other authors have no competing interests to declare. Supplementary Files TableS1.xlsx Supplementary Table 1 TableS2.xlsx Supplementary Table 2 TableS3.xlsx Supplementary Table 3 TableS4.xlsx Supplementary Table 4 TableS5.xlsx Supplementary Table 5 TableS6.xlsx Supplementary Table 6 TableS7.xlsx Supplementary Table 7 TableS8.xlsx Supplementary Table 8 TableS9.xlsx Supplementary Table 9 Supplementaryinformation.docx Supplementary Information ExtendFigures.pdf Extended Data Figure Legends Extended Data Fig. 1. Human liver cell types display unique patterns of chromatin accessibility. (A-F) Quality control metrics for snATAC-seq data. Shown are the number of captured nuclei per sample (A), read distribution across genomic regions (B), transcription start site (TSS) enrichment scores (C), nucleosome signal profiles (D), percentage of reads mapping to called peaks (E), and relative cell type proportions across samples (F). (G) Genomic annotation of accessible peaks assigned to each major liver cell type. (H) UMAP visualization of fully integrated snATAC-seq data from normal and MASLD liver samples. (I) UMAP embedding visualized with coloration distinguishing between normal and MASLD individuals. (J) Comparison of cell type distributions between normal and MASLD liver tissues. (K) Genome browser tracks showing aggregated chromatin accessibility at marker gene loci across major liver cell types. Data are presented as mean ± SEM. Statistical significance was assessed using Student’s t-test. Extended Data Fig. 2. Human liver cell types exhibit distinct transcriptional regulatory program. (A) Heatmap of cell type–specific chromatin accessibility peaks. Colors represent normalized marker peak intensities across major liver cell types. (B-C) Heatmap (B) and UMAP visualization (C) of enriched transcription factor (TF) motifs across each cell type. Colors indicate normalized motif enrichment scores. (D-E) Integrated analysis of our snATAC-seq data with published human liver snATAC-seq profiles from Kim et al. UMAP projection of merged cell types across both datasets (D). Correlation matrix comparing cell type annotations between the two studies (E). (F) Heatmap of cell type-specific marker gene activity derived from snATAC-seq gene scores. Colors denote normalized accessibility-derived gene scores. (G) UMAP plot displaying cell type-specific marker gene scores for KRT7 (cholangiocytes), FLT1 (endothelial cells), ALB (hepatocytes), CD163 (Kupffer cells), CD8A (natural killer cells), and NGF (stellate cells). (H) Heatmap illustrating alignment between cell type annotations from snATAC-seq and snRNA-seq datasets from Xiao et al. Extended Data Fig. 3. Enrichment of MASLD-associated variants in hepatocyte and stellate cells. (A-B) Enrichment of GWAS variants associated with plasma triglycerides, body mass index (BMI), and MASLD in cell type-specific accessible chromatin regions. Regression coefficients (A) and enrichment scores (B) were calculated using linkage disequilibrium score regression (LDSC). (C-D) Gene ontology (GO) enrichment analysis of MASLD-specific and normal-specific stellate cell subclusters identified in Figure 1F. (E-F) Pseudotime trajectory analysis of stellate cells inferred by StaVia (E), and correlation of StaVia pseudotime values with those derived from Monocle3 (F), demonstrating consistency across trajectory inference methods. (G-H) GO enrichment analysis of MASLD-specific and normal-specific hepatocyte subclusters shown in Figure 1L. (I-J) Pseudotime analysis of hepatocyte populations using StaVia (I) and comparison with Monocle3-based pseudotime estimates (J), supporting robustness of inferred dynamic transitions. Extended Data Fig. 4. Enrichment analysis of liver disease-associated variants within relevant liver cell types. (A-B) UMAP projection of three transcriptionally distinct Kupffer cell subclusters (A) and their relative proportions across normal and MASLD liver tissues (B). (C-D) Gene ontology (GO) enrichment analysis of normal-specific and MASLD-specific Kupffer cell clusters. (E-F) Trait Relevance Score (TRS) analysis of Kupffer cell subclusters for MASLD GWAS traits. UMAP visualization of SCAVENGE-inferred TRSs (E). Box plots showing TRS distributions across the three subclusters (F). Boxes represent the interquartile range (IQR), horizontal lines indicate the median, and whiskers denote 1.5× IQR. (G) Correlation between motif enrichment scores and MASLD GWAS variant enrichment within hepatocytes. Top 5 motifs were shown. (H) UMAP visualization of SCAVENGE TRSs for the cirrhosis trait across all liver cell types. (I) Correlation between TF motif enrichment and cirrhosis GWAS variant enrichment within hepatocytes. Top five enriched motifs are shown. (J) Hepatocytes-specific gene regulatory networks (GRN) in Normal and MASLD condition. Each node in the network represents a gene or transcription factor, with its color and size denoting the expression level and PageRank centrality of the respective gene or transcription factor. The edges between nodes indicate the statistical significance of interactions between transcription factors and genes. Only the nodes identified as transcription factors are labeled in the network. (K) The expression levels and motif activities (represented as chromVAR deviation Z-scores) of transcription factors between the GRNs of MASLD and Normal hepatocytes in extended data figure 4J. Extended Data Fig. 5. Design of the MPRA and assessment of its reproducibility. (A) Schematic diagram outlining the strategy for selecting MASLD-associated genetic variants for inclusion in the MPRA library. (B) Intersection of total MPRA variants with liver cell type-specific promoter and enhancer regions. Cell type-specific promoter regions defined herein refer to promoter regions (2 kb upstream and 0.5 kb downstream of the TSS) that overlap with cell type-specific peaks, while the remaining cell type-specific peaks are designated as cell type-specific enhancers. (C-D) Odds ratios (C) and (D) significance levels demonstrating the enrichment of total MPRA variants in liver cell type-specific promoter and enhancer regions versus random genomic regions. (E) Histogram depicting the complexity of the MPRA library, featuring a median of 714 barcodes assigned per element. (F) GFP fluorescence intensities demonstrate higher enhancer activity for positive CRS, moderate activity for MPRA variants, and limited activity for negative CRS in HepG2 cells. (G) Quantitative MPRA activity across variant sets, including positive controls, negative controls, and both alleles of tested MPRA variants, measured in HepG2 cells. (H) Reproducibility analysis of MPRA performed in HepG2 cells across four biological replicates, showing high correlation and technical robustness. Extended Data Fig. 6. Characterization of DAVs identified by MPRA in HepG2. (A) TF motif enrichment analysis of active genomic elements (at least one active allele) in HepG2 cells, using other inactive element as background. (B) Manhattan plot of 5,369 MASLD variants assessed by MPRA. Red dots indicate genome-wide significant variants ( p < 1 × 10⁻ 6 ). (C) Genomic distribution of all tested MPRA variants across chromosomes. DAVs (FDR < 0.01) are highlighted in red. (D) Pie chart showing the distribution of DAVs across different genomic categories (e.g., promoter, intronic, intergenic, enhancer regions). (E) Pie chart illustrating the number of functional genomic annotations associated with each DAV, including ATAC peaks from snATAC-seq data, liver eQTLs, chromatin interaction loci, disrupted TF binding motifs, and ChromHMM states. (F-G) Overlap of DAVs with liver cell type-specific chromatin accessibility peaks (F) and total peaks across all liver cell types (G). (H) Enrichment analysis showing the overrepresentation of DAVs in cell type-specific accessible chromatin regions relative to all tested MPRA variants. Enrichment is represented by -log10( p ) and visualized by color scale. (I) The enrichment of transcription factor motifs affected by DAVs, as identified by the motifbreakR algorithm. Extended Data Fig. 7. Visualizations of DAVs identified by MPRA in HepG2. (A) Genomic browser view of the GLDC locus. DAVs (highlighted in yellow) are shown within hepatocyte-specific accessible chromatin peaks, annotated enhancer regions, and chromatin loop anchors. (B) Genomic browser view of the MLXIP locus. DAVs (highlighted in yellow) are located within hepatocyte-specific accessible chromatin peaks, annotated enhancer regions, and chromatin loop anchors. Extended Data Fig. 8. Motif activities of transcription factors influenced by DAVs in normal versus MASLD livers. (A) UMAP visualization of DAVs-affected TF motifs across each cell type, including GABPA, HNF4G, KLF5, MYC, SP1, SP2, SREBF1, and SREBF2. Colors indicate normalized motif enrichment scores. (B) Box plots showing the motif activity of DAVs-affected TFs in normal versus MASLD livers. Boxplots show median as a horizontal line, interquartile range as a box. (C-E) UMAP visualization of DAVs-affected TF motifs within hepatocytes, including GABPA (C), CREB3 (D), and KLF15 (E). Colors indicate normalized motif enrichment scores. Box plots showing the motif activity of these DAVs-affected TFs in normal versus MASLD hepatocytes. Boxplots show median as a horizontal line, interquartile range as a box. Extended Data Fig. 9. Identification and characterization of cell type-specific DAVs. (A) Volcano plot illustrating the activity of cell type-specific elements. Elements with significantly differential activity (BH corrected p value < 0.05) are highlighted in red and blue dots. (B) TF motif enrichment analysis of LX-2-specific regulatory elements, indicating enrichment for fibrosis-associated TFs. (C) Enhancer activity for each element was plotted according to its activity rank in LX-2. Alleles were classified as Enhancers (red) if they had an adjusted p value 0. (D) Intersection sizes of LX-2 DAVs overlapping with functional annotations such as ATAC peaks from snATAC-seq data, liver eQTLs, liver chromatin interactions, transcription factor binding motifs disruption, and ChromHMM states. (E) Pie chart showing the number of functional annotations associated with each LX-2 DAV. (F) TF motif enrichment analysis of cell type-specific DAVs. Colors indicate significantly enriched motifs (q < 0.1). Extended Data Fig. 10. Identification and characterization of stimuli-induced DAVs. (A) Volcano plot illustrating the activity of PAOA-induced elements. Elements with significantly differential activity (BH corrected p value < 0.05) are highlighted in red and blue dots. (B-C) TF motif enrichment analysis of PAOA-responsive regulatory elements. Red and blue colors denote significantly enriched motifs (q < 0.1). The FOXF1, BMAL1, HNF6 motifs highlighted in light blue and THRβ motif highlighted in light Salmon color indicate nominal significance (p < 0.01). (D) Enhancer activity for each element was plotted according to its activity rank in PAOA-treated HepG2. Alleles were classified as Enhancers (red) if they had an adjusted p -value 0. (E) The enrichment of transcription factor motifs affected by PAOA-HepG2 DAVs, as identified by the SNP2TFBS algorithm. (F) Scatter plots illustrating the allele-differential activities of MPRA variants between PAOA- and Ctrl-treated conditions. (G) Enhancer activity for each element was plotted according to its activity rank in TGFb-treated LX-2 cell. Alleles were classified as Enhancers (red) if they had an adjusted p -value 0. (H) Intersection sizes of TGFb-treated LX-2 DAVs overlapping with functional annotations such as ATAC peaks from snATAC-seq data, liver eQTLs, liver chromatin interactions, transcription factor binding motifs disruption, and ChromHMM states. (I) The enrichment of transcription factor motifs affected by TGFb-treated LX-2 DAVs, as identified by the SNP2TFBS algorithm. Extended Data Fig. 11. DAVs-controlled transcriptional network. (A) Protein-protein interaction network of all DAVs disrupted TFs from four MPRA datasets. (B) WikiPathways enrichment of all DAVs disrupted TFs from four MPRA datasets. Extended Data Fig. 12. Determination of target genes associated with MASLD DAVs. (A) Venn diagram illustrating the overlap of DAVs across different conditions. (B) Comparison of target gene numbers identified by the four approaches. (C-F) Distribution of the number of target genes assigned to each DAV in HepG2 (C), PAOA-treated HepG2 (D), LX-2 (E), and TGFb-treated LX-2 (F). (G-J) Pie chart illustrating the number of DAVs overlapping with liver eQTLs that exhibit concordant or discordant allele-specific MPRA activity relative to eQTL effect sizes in HepG2 (G), PAOA-treated HepG2 (H), LX-2 (I), and TGFb-treated LX-2 (J). (K) GRN of DAVs target genes. Each node in the network represents a gene or transcription factor, with its color and size denoting the log2 fold change of expression level between MASLD and Normal, and PageRank centrality of the respective gene or transcription factor. The edges between nodes indicate the statistical significance of interactions between transcription factors and genes. Only the nodes identified as transcription factors are labeled in the network. Extended Data Fig. 13. CROP-seq of candidate DAVs. (A) Immunofluorescence images displaying Cas9 expression in dCas9-KRAB/MeCP2 and dCas9-VPR knockin HepG2 cell line. Scale bar: 200μm. (B) mRNA expression of nominated target genes in HepG2 cell following introduction of CRISPRa machinery (n = 6). (C) The workflow outlining the selection criteria of CROP-seq experiment. (D) Distribution of the number of cells targeted by each individual sgRNA in the CROP-seq screen. (E) Total number of cells associated with each candidate DAV-targeting sgRNA across the dataset. (F) UMAP visualization of single-cell transcriptomic profiles from the CROP-seq experiment in HepG2 cells infected with lentiviral sgRNA constructs targeting DAV-associated regulatory regions. Data are expressed as mean ± SEM (Student’s t test). Extended Data Fig. 14. Functional detection of DAVs in HepG2 cell line and LX-2 cell lines. (A) Agarose gel electrophoresis of PCR-amplified genomic regions in stable DAV-deleted HepG2 or LX-2 cell lines. Genomic loci corresponding to rs1748197, rs10889356, rs10790167, rs1441754, and rs66477683 were used as PCR templates. Bands highlighted in black rectangles indicate successful deletion of the corresponding DAV region. (B) Position weight matrix (PWM) of the MYC binding motif (minus strand) predicted to be disrupted by rs1748197. (C) mRNA expression of ANGPTL3 and USP1 in HepG2 cells following CRISPR-mediated deletion (Del), CRISPR interference (CRISPRi, CRi), or CRISPR activation (CRISPRa, CRa) (n = 3-6). (D) PWM of the GABPA binding motif (minus strand) predicted to be altered by rs10889356. (E) Luciferase reporter activity of the two alleles of rs10889356 in 293T cells expressing pLVX-Ctrl or pLVX-GABPA. (F) Flow cytometry analysis of lipid accumulation in rs10889356-deleted HepG2 cells treated with or without PAOA for 24 h (n = 3). (G) Representative images of BODIPY staining in rs10889356 deleted-HepG2 treated with PAOA for 24 h. Scale bars, 10 µm. (H) Representative images of BODIPY staining in 3D Liver Spheroids treated with PAOA for 48 h. Scale bars, 100 µm. (I) PWMs of RXR and PPARα binding motifs (minus strand) affected by rs10790167. (J) Luciferase reporter activity of rs1441754 alleles across three liver cell lines: LX-2, HepG2, and Huh7. (K) The putative effect of rs1441754 on RUNX binding. Data are expressed as mean ± SEM (Student’s t test). Extended Data Fig. 15. Functional detection of rs66477683 in LX-2 cell line. (A) Genomic browser view of the rs66477683 locus (highlighted in yellow) near the MERTK gene, annotated with regulatory features including chromatin accessibility, enhancer elements, and chromatin interactions. (B) Luciferase reporter activity of the two alleles of rs66477683 in 293T cells, assessing allele-specific regulatory function. (C) mRNA expression of MERTK and fibrosis-related genes, including TIMP1 , COL1A1 and ACTA2 in rs66477683-deleted LX-2 (n = 3). (D) Representative immunofluorescence images of COL1A1 and ACTA2 in LX-2 cells treated with or without 20 ng/mL TGFβ for 48 h. Scale bars, 25 µm. Data are expressed as mean ± SEM (Student’s t test). Extended Data Fig. 16. Functional detection of rs474513 in HepG2, Huh7 cell lines and iPSCs-derived hepatic-like cells. (A) Agarose gel electrophoresis of PCR-amplified genomic DNA from rs474513-deleted HepG2 cells. Bands highlighted in black rectangles indicate successful deletion of the DAV-containing region. (B) mRNA expression of SLC22A3 in rs474513-deleted HepG2 cells treated with the RUNX2 inhibitor CADD522 (n = 4), as quantified by qRT-PCR and normalized to HPRT. (C) mRNA expression of SLC22A3 and LPA in HepG2 cells following CRISPR interference (CRISPRi)-mediated repression (n = 3-6). (D) Sanger sequencing chromatogram confirming correction of the G/G genotype to G/A at rs474513 in edited cells. (E) SLC22A3 mRNA expression in HepG2 cells transfected with SLC22A3 overexpression plasmid for 24 h, followed by treatment with RUNX inhibitors (RO5-3335 for RUNX1 and CADD522 for RUNX2) for another 24 h. (F) Quantification of BODIPY intensity per cell from Figure 7K using ImageJ (n = 3 fields). (G) Agarose gel electrophoresis of PCR products from rs474513-deleted iPSC lines. Bands within black rectangles denote successful deletion. (H) Relative mRNA expression of SLC22A3 in rs474513-deleted iPSC lines. (I) Relative mRNA expression of marker genes at different stages of hepatic-like cell differentiation derived from iPSCs. (J) Comparison of odds ratios for PRS constructed from 114 LD-independent DAVs for MASLD traits in the All of Us cohort (weighted by MVP GWAS, Vujkovic et al) versus PRS9, created from 9 Missense SNPs (Methods), and published PRS (PGS002732 and PGS000655) from the PGS Catalog. Data are expressed as mean ± SEM (Student’s t test). Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6984670","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":530320156,"identity":"1c4212e1-5f14-4610-a2e7-faec53ae2e09","order_by":0,"name":"Biying Zhu","email":"","orcid":"","institution":"Guangzhou National Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Biying","middleName":"","lastName":"Zhu","suffix":""},{"id":530320157,"identity":"f0175376-2f22-4f41-8edc-838c0bced9e3","order_by":1,"name":"Na He","email":"","orcid":"","institution":"Guangzhou National Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Na","middleName":"","lastName":"He","suffix":""},{"id":530320158,"identity":"715541e1-81f5-4a53-a1d5-24aa2e87a156","order_by":2,"name":"Yang Xiao","email":"","orcid":"","institution":"University of Pennsylvania","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Xiao","suffix":""},{"id":530320159,"identity":"1e4c684d-cc32-4e1f-889d-a2cf1c100e2e","order_by":3,"name":"Bin Chen","email":"","orcid":"","institution":"Guangzhou National Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Bin","middleName":"","lastName":"Chen","suffix":""},{"id":530320160,"identity":"6a4d131f-5d8d-4d0e-a4b0-bb49581dd8b0","order_by":4,"name":"Chen Li","email":"","orcid":"","institution":"Guangzhou National Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Chen","middleName":"","lastName":"Li","suffix":""},{"id":530320161,"identity":"0d7a3dd5-fc63-4f4b-b974-ed8130a812e0","order_by":5,"name":"Ravi Mandla","email":"","orcid":"https://orcid.org/0000-0002-0782-0138","institution":"University of Pennsylvania","correspondingAuthor":false,"prefix":"","firstName":"Ravi","middleName":"","lastName":"Mandla","suffix":""},{"id":530320162,"identity":"1430d420-a1db-4426-8ac0-d8648cd44899","order_by":6,"name":"Yifan Liu","email":"","orcid":"","institution":"University of Pennsylvania","correspondingAuthor":false,"prefix":"","firstName":"Yifan","middleName":"","lastName":"Liu","suffix":""},{"id":530320163,"identity":"ac06e781-6f2a-46a0-aa19-85e029750e1c","order_by":7,"name":"Jiayu Zhang","email":"","orcid":"","institution":"University of Pennsylvania","correspondingAuthor":false,"prefix":"","firstName":"Jiayu","middleName":"","lastName":"Zhang","suffix":""},{"id":530320164,"identity":"fa904d30-81b2-4036-9d2b-1171a7b8e6c7","order_by":8,"name":"Xiao Chang","email":"","orcid":"","institution":"Shandong First Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xiao","middleName":"","lastName":"Chang","suffix":""},{"id":530320165,"identity":"003fda52-4140-46ec-9a5d-1a7808e11a4f","order_by":9,"name":"Fulong Yu","email":"","orcid":"","institution":"Guangzhou National Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Fulong","middleName":"","lastName":"Yu","suffix":""},{"id":530320166,"identity":"f321f8a7-1536-46e3-a684-e7e3417e0ab5","order_by":10,"name":"Marijana Vujkovic","email":"","orcid":"https://orcid.org/0000-0003-4924-5714","institution":"University of Pennsylvania School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Marijana","middleName":"","lastName":"Vujkovic","suffix":""},{"id":530320167,"identity":"011e4bcf-4f7d-4d23-8e86-b32b3cb6ea2a","order_by":11,"name":"Julie Lynch","email":"","orcid":"https://orcid.org/0000-0003-0108-2127","institution":"Salt Lake City VA Healthcare System","correspondingAuthor":false,"prefix":"","firstName":"Julie","middleName":"","lastName":"Lynch","suffix":""},{"id":530320168,"identity":"2eb5963d-d558-43d5-bad5-305453f3f504","order_by":12,"name":"Kyong-Mi Chang","email":"","orcid":"https://orcid.org/0000-0001-6811-9364","institution":"The Corporal Michael J. Crescenz Veterans Affairs Medical Center and University of Pennsylvania Perelman School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Kyong-Mi","middleName":"","lastName":"Chang","suffix":""},{"id":530320169,"identity":"69d5ed67-12aa-4229-b9db-be5eac8918f6","order_by":13,"name":"VA Million Veteran Program","email":"","orcid":"","institution":"U.S Department of Veteran Affairs","correspondingAuthor":false,"prefix":"","firstName":"VA","middleName":"Million Veteran","lastName":"Program","suffix":""},{"id":530320170,"identity":"0e7a1be3-878a-47c6-bd7b-9c6587de3530","order_by":14,"name":"Bogdan Pasaniuc","email":"","orcid":"","institution":"University of Pennsylvania","correspondingAuthor":false,"prefix":"","firstName":"Bogdan","middleName":"","lastName":"Pasaniuc","suffix":""},{"id":530320171,"identity":"988ab493-03b8-4a5e-9765-69d635500c3d","order_by":15,"name":"Daniel Rader","email":"","orcid":"https://orcid.org/0000-0002-9245-9876","institution":"University of Pennsylvania","correspondingAuthor":false,"prefix":"","firstName":"Daniel","middleName":"","lastName":"Rader","suffix":""},{"id":530320172,"identity":"32e0fd81-7c94-4fd3-9052-3796c9e8799c","order_by":16,"name":"Mitchell A. Lazar","email":"","orcid":"https://orcid.org/0000-0001-8653-1280","institution":"University of Pennsylvania","correspondingAuthor":false,"prefix":"","firstName":"Mitchell","middleName":"A.","lastName":"Lazar","suffix":""},{"id":530320155,"identity":"1e7df417-1b36-4e73-9985-5b09643fd999","order_by":17,"name":"Wenxiang Hu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABAElEQVRIiWNgGAWjYBACAyjNA8SMDxgYLOAiRGlhBnIkiNcCAmwSRGkxl0hgk/i447AM/+z2a5Vf/kjImzMwP/zAUHMHpxbLGQlskjPPpPFI3DlTdluGR8JwZwObsQTDsWe4HXYjge02b5sND8ONnLTbEhISjBsOMJgxMDYcxq/lb5sEjzxQS7GEgYT9hgPs3whrYQTaYnAj/RjjhwSJxA0HeAjYcuYB+8/etjQewxs5zNIMBySSdzbzFEskHMOj5XgCs8HPtsP2cjfSH3788cfGdjt7+8YPH2pwa2Fg4P8AZfAYMIMilIEZiBPwaEAC7A8YfxCnchSMglEwCkYYAAAOcVKqAWcxGAAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0002-6754-5625","institution":"Guangzhou National Laboratory","correspondingAuthor":true,"prefix":"","firstName":"Wenxiang","middleName":"","lastName":"Hu","suffix":""}],"badges":[],"createdAt":"2025-06-26 14:56:43","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6984670/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6984670/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":94162114,"identity":"2fa20f72-723a-4d73-87be-06a35ebe82d5","added_by":"auto","created_at":"2025-10-23 05:06:51","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":397790,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscript.docx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/833cb96e28fee5b428d90f55.docx"},{"id":94161631,"identity":"3b1321e5-d78e-4ddc-85ed-02429a949f84","added_by":"auto","created_at":"2025-10-23 04:58:51","extension":"json","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":16623,"visible":true,"origin":"","legend":"","description":"","filename":"NGA69532R.json","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/9f7960cdae80d74f8d8e83cd.json"},{"id":94161637,"identity":"47e7b2a2-f830-4e25-9a1d-48e832c8e75c","added_by":"auto","created_at":"2025-10-23 04:58:51","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":150933,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaryinformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/ac1ac1590544fc857b6d2000.docx"},{"id":94161642,"identity":"542ca881-9d76-45f4-be80-8d6e8e9113bc","added_by":"auto","created_at":"2025-10-23 04:58:51","extension":"xlsx","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15656,"visible":true,"origin":"","legend":"","description":"","filename":"TableS1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/4ab56780191d2203139b4793.xlsx"},{"id":94161636,"identity":"f761f2af-0804-4466-aac9-272ca1ba8e0f","added_by":"auto","created_at":"2025-10-23 04:58:51","extension":"xlsx","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39222,"visible":true,"origin":"","legend":"","description":"","filename":"TableS2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/22bc96130c18cf29698dcd13.xlsx"},{"id":94162595,"identity":"9b8cddbc-d2da-4351-84d2-b2436d99d3e3","added_by":"auto","created_at":"2025-10-23 05:14:51","extension":"xlsx","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":46645,"visible":true,"origin":"","legend":"","description":"","filename":"TableS3.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/93dc467c6a52dc386179ba48.xlsx"},{"id":94162116,"identity":"2682e520-372b-4ec8-bcbb-b8fd28b77e73","added_by":"auto","created_at":"2025-10-23 05:06:51","extension":"xlsx","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":37153,"visible":true,"origin":"","legend":"","description":"","filename":"TableS4.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/e0da7b5e004203bc7a583211.xlsx"},{"id":94161641,"identity":"6780bc6c-513e-45ac-a09a-a47576ac7755","added_by":"auto","created_at":"2025-10-23 04:58:51","extension":"xlsx","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":58929,"visible":true,"origin":"","legend":"","description":"","filename":"TableS5.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/ae3be91c934914caef683357.xlsx"},{"id":94162120,"identity":"231b214b-c0ba-4dcc-bd08-1a29b3d35745","added_by":"auto","created_at":"2025-10-23 05:06:52","extension":"xlsx","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15053,"visible":true,"origin":"","legend":"","description":"","filename":"TableS6.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/30ae39e08ead748ae3030636.xlsx"},{"id":94162119,"identity":"67dd4d9d-82c9-4772-9966-aaf089e93381","added_by":"auto","created_at":"2025-10-23 05:06:51","extension":"xlsx","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":24300,"visible":true,"origin":"","legend":"","description":"","filename":"TableS7.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/2d4a76cba2feb7a1957c96fa.xlsx"},{"id":94161657,"identity":"6c4633d8-b885-445b-923f-78e2f303f836","added_by":"auto","created_at":"2025-10-23 04:58:52","extension":"xlsx","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14804,"visible":true,"origin":"","legend":"","description":"","filename":"TableS8.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/1856508fe3c08fde3421afb8.xlsx"},{"id":94161644,"identity":"c3e49018-3265-4ec0-8279-2c716a1f02bd","added_by":"auto","created_at":"2025-10-23 04:58:51","extension":"xlsx","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":31849,"visible":true,"origin":"","legend":"","description":"","filename":"TableS9.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/fe6978d93fd08b31460edf2c.xlsx"},{"id":94162123,"identity":"5f99ab29-08ae-41a1-8698-46330c0fcec0","added_by":"auto","created_at":"2025-10-23 05:06:52","extension":"xml","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":298062,"visible":true,"origin":"","legend":"","description":"","filename":"NGA69532R1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/3740f55765785076c954b624.xml"},{"id":94161655,"identity":"66e8d999-342a-429d-865c-024f50b34b4c","added_by":"auto","created_at":"2025-10-23 04:58:52","extension":"pdf","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11876195,"visible":true,"origin":"","legend":"","description":"","filename":"FiguresandsFigures.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/a2d6951a0244c06be171a76e.pdf"},{"id":94161650,"identity":"557a1a43-4f39-4672-87e7-50d6594915f9","added_by":"auto","created_at":"2025-10-23 04:58:52","extension":"xml","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":293958,"visible":true,"origin":"","legend":"","description":"","filename":"NGA69532R1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/f34f87e6673570260c8015a1.xml"},{"id":94161651,"identity":"e3b3f166-fdf5-4c96-bf92-f088523fdc96","added_by":"auto","created_at":"2025-10-23 04:58:52","extension":"html","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":329173,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/9b32bfdbe7e2bae29e7daa9c.html"},{"id":94161626,"identity":"24193ef8-9945-461f-94c7-fb3535a76540","added_by":"auto","created_at":"2025-10-23 04:58:51","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1749142,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCell type-specific accessible chromatin in the liver enriches for MASLD risk variants.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e Schematic illustration of snATAC-seq in normal and MASLD livers.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B)\u003c/strong\u003e Fully integrated UMAP of 69,595 nuclei from human livers identifies 15 distinct clusters labeled 0-14.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C)\u003c/strong\u003e Comparison of cell type proportions between normal and MASLD liver samples. Bar plots show the relative abundance of each annotated cell population under the two conditions (n = 6).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(D-E)\u003c/strong\u003e UMAP visualization of per-cell Trait Relevance Scores (TRSs) calculated by SCAVENGE for the liver fat trait (D) and MASLD trait (E). Cells are colored according to their TRS, highlighting trait-associated cell populations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(F-G)\u003c/strong\u003e UMAP visualization of three distinct hepatic stellate cell subclusters (F) and their relative proportions across normal and MASLD conditions (G).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(H)\u003c/strong\u003e Box plots illustrating SCAVENGE TRS scores for MASLD traits across the three stellate cell subclusters. Boxes represent the interquartile range (IQR), horizontal lines indicate the median, and whiskers denote 1.5× IQR.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(I)\u003c/strong\u003e Pseudotime trajectory overlaid on the UMAP of hepatic stellate cells, illustrating the inferred dynamic transition states among subclusters.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(J-K)\u003c/strong\u003e UMAP (J) and box plots (K) illustratingSCAVENGE TRS scores for MASLD traits across the stellate cell activation trajectory. Boxes represent the interquartile range (IQR), horizontal lines indicate the median, and whiskers denote 1.5× IQR.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(L-M)\u003c/strong\u003e UMAP visualization of four distinct hepatocyte subclusters (L) and their relative proportions across normal and MASLD conditions (M).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(N-O)\u003c/strong\u003e UMAP (N) and box plots (O) illustrating SCAVENGE TRS scores for MASLD traits along the hepatic remodeling trajectory.\u003cstrong\u003e \u003c/strong\u003eBoxes represent the interquartile range (IQR), horizontal lines indicate the median, and whiskers denote 1.5× IQR.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(P-Q)\u003c/strong\u003e Correlation of TF motif enrichment with MASLD- (P) and Cirrhosis-associated risk variants (Q) enrichment in all cells. TF motif activity for each cell was quantified using chromVAR, and TRS for individual cells were computed using SCAVENGE. Spearman’s rank correlation coefficients were then calculated between motif activity and TRS across all cells to assess the association between transcription factor activity and GWAS trait enrichment.Top 5 motifs were shown.\u003c/p\u003e\n\u003cp\u003eData are presented as mean ± SEM for proportion analyses shown in panels (C), (G), and (M). Statistical significance was determined using two-tailed Student’s t-test.\u003c/p\u003e","description":"","filename":"FiguresandsFigures1.png","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/f9ce81dbc113963e1c187e0a.png"},{"id":94161627,"identity":"baf7b545-c5b3-41ca-8af6-34f0d4dda050","added_by":"auto","created_at":"2025-10-23 04:58:51","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1322850,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eIdentification and characterization of functional variants using MPRA.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e Schematic overview of the MPRA experimental workflow. The assay was performed in four biological replicates per condition to ensure reproducibility of enhancer activity measurements.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B) \u003c/strong\u003eEnhancer activity for each element was plotted according to its activity rank. Alleles were classified as Enhancers (red) if they had an adjusted \u003cem\u003ep\u003c/em\u003e-value \u0026lt; 0.05 and a log2 fold change \u0026gt; 0. Active elements were transcriptionally active for at least one allele.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C) \u003c/strong\u003eVolcano plot showing the effect sizes and \u003cem\u003ep\u003c/em\u003e values of MPRA variants. Variants with significant (FDR \u0026lt; 0.01) allele-differential activity (DAVs) are highlighted in red dots.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(D)\u003c/strong\u003e The activities of luciferase reporters with the different alleles for ten random DAVs\u003cstrong\u003e \u003c/strong\u003ein HepG2. Data are shown as mean ± SEM fold change relative to their own ref allele (Student’s t test).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(E) \u003c/strong\u003eIntersection sizes of DAVs overlapping with functional annotations such as ATAC peaks from snATAC-seq data, liver eQTLs, liver chromatin interactions, transcription factor binding motif disruption, and ChromHMM states.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(F) \u003c/strong\u003eEnrichment of DAVs within liver cell type-specific accessible chromatin regions. Total tested variants were used as the background set. \u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eThe \u003cem\u003ep\u003c/em\u003e values were showed (Fisher’s exact test).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(G) \u003c/strong\u003eCorrelation between DAV effect sizes and TF motif disruption scores, indicating functional consequences of allelic variants on TF binding.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(H) \u003c/strong\u003eThe enrichment of transcription factor motifs affected by DAVs, as identified by the SNP2TFBS algorithm.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(I) \u003c/strong\u003eUMAP plots displaying the motif activity of DAV-affected TFs, including HNF4G and SREBF2, within normal and MASLD hepatocytes. Colors indicate normalized motif enrichment scores.\u003c/p\u003e","description":"","filename":"FiguresandsFigures2.png","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/e0a7690efae7f5dd51ed1c1c.png"},{"id":94162112,"identity":"672ca7d7-8618-4fa7-8b55-adc2f2e18889","added_by":"auto","created_at":"2025-10-23 05:06:51","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1556919,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eIdentification of cell type-specific DAVs.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e Heatmap of cell type-specific active elements significantly active in HepG2 or LX-2 cell lines. Differential element activity were identified with FDR \u0026lt; 0.05. Color scale represents the ratio of RNA/DNA in normalized counts.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B) \u003c/strong\u003eTF\u003cstrong\u003e \u003c/strong\u003emotif enrichment analysis for cell type-specific active elements identified in HepG2 and LX-2. Colors indicate significantly enriched motifs (q \u0026lt; 0.1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C) \u003c/strong\u003eVolcano plot showing the effect sizes and \u003cem\u003ep\u003c/em\u003e values of MPRA variants identified in LX-2. Variants with significant (FDR \u0026lt; 0.01) allele-differential activity (DAVs) are highlighted in red dots.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(D) \u003c/strong\u003eVenn diagram demonstrating the overlapped DAVs in HepG2 and LX-2 cell lines.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(E) \u003c/strong\u003eScatter plots illustrating the allele-differential activities of DAVs between HepG2 and LX-2 cell lines.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(F)\u003c/strong\u003e TF motif enrichment analysis of DAVs specific to LX-2 cells, highlighting enrichment of fibrosis-related TFs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(G) \u003c/strong\u003eMPRA-assessed activity of the A and G alleles of rs6726823 in HepG2 and LX-2 cell lines.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(H)\u003c/strong\u003e Genomic visualization of the rs6726823 locus near \u003cem\u003eCASP8\u003c/em\u003e. The position of rs6726823 is highlighted in yellow. Annotations include snATAC chromatin accessibility, liver histone modification profiles, chromatin loops and regulatory links, MPRA-tested regions, and identified DAVs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(I) \u003c/strong\u003ePredicted effect of rs6726823 on IRF binding affinity, based on TF motif disruption analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(J)\u003c/strong\u003e MPRA-assessed activity of the A and C alleles of rs13150068 in HepG2 and LX-2 cell lines.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(K) \u003c/strong\u003eGenomic visualization of rs13150068 within hepatocyte-specific peak (yellow box) at \u003cem\u003eHSD17B13 \u003c/em\u003elocus with relevant regulatory annotations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(L) \u003c/strong\u003ePredicted effect of rs13150068 on SREBP2 binding affinity, based on TF motif disruption analysis.\u003c/p\u003e","description":"","filename":"FiguresandsFigures3.png","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/e48c60e9e5598f33ef4373f2.png"},{"id":94162594,"identity":"4e3050e0-da1a-417e-9c53-cc479e194043","added_by":"auto","created_at":"2025-10-23 05:14:51","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1507011,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eContext-dependent activity of DAVs\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A) \u003c/strong\u003eVolcano plot showing the effect sizes and \u003cem\u003ep\u003c/em\u003e values of MPRA variants identified in PAOA-treated HepG2 cell. Variants with significant (FDR \u0026lt; 0.01) allele-differential activity (DAVs) are highlighted in red dots.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B)\u003c/strong\u003e Intersection sizes of PAOA-HepG2 DAVs overlapping with functional annotations such as ATAC peaks from snATAC-seq data, liver eQTLs, liver chromatin interactions, transcription factor binding motifs disruption, and ChromHMM states.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C) \u003c/strong\u003eVenn diagram demonstrating the overlapped DAVs identified in Ctrl- and PAOA-treated HepG2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(D) \u003c/strong\u003eTF\u003cstrong\u003e \u003c/strong\u003emotif enrichment analysis of stimuli-specific DAVs.\u003cstrong\u003e \u003c/strong\u003eRed and blue colors denote significantly enriched motifs (q \u0026lt; 0.1). The PPARE motif is highlighted in light salmon to indicate nominal significance (p \u0026lt;0.01).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(E) \u003c/strong\u003eMPRA-assessed activity of the T and G alleles of rs12406250 in Ctrl- and PAOA-treated HepG2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(F) \u003c/strong\u003eGenomic visualization of rs12406250 within MASLD-enhanced hepatocyte peak (yellow box) at \u003cem\u003eEPHA2\u003c/em\u003e locus. The position of rs12406250 is highlighted in yellow. Annotations include snATAC chromatin accessibility, liver histone modification profiles, chromatin loops and regulatory links, MPRA-tested regions, and identified DAVs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(G)\u003c/strong\u003eVolcano plot showing the effect sizes and \u003cem\u003ep\u003c/em\u003e values of MPRA variants identified in TGFb-treated LX-2 cell. Variants with significant (FDR \u0026lt; 0.01) allele-differential activity (DAVs) are highlighted in red dots.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(H) \u003c/strong\u003eVenn diagram demonstrating the overlapped DAVs in Ctrl- and TGFb-treated LX-2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(I)\u003c/strong\u003e Scatter plots illustrating the allele-differential activities of MPRA variants between TGFb- and Ctrl-treated conditions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(J) \u003c/strong\u003eTF\u003cstrong\u003e \u003c/strong\u003emotif enrichment analysis of TGFb stimuli-specific DAVs.\u003cstrong\u003e \u003c/strong\u003eColors indicate significantly enriched motifs (q \u0026lt; 0.1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(K) \u003c/strong\u003eMPRA-assessed activity of the A and G alleles of rs558702 in Ctrl- and TGFb-treated LX-2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(L) \u003c/strong\u003eGenomic visualization of rs558702 (yellow box) at \u003cem\u003eZBTB12\u003c/em\u003elocus with relevant regulatory annotations.\u003c/p\u003e","description":"","filename":"FiguresandsFigures4.png","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/43fd7af474e19c90de0d996c.png"},{"id":94161653,"identity":"feec63c0-9442-4be6-b1f4-e78ed29f40a9","added_by":"auto","created_at":"2025-10-23 04:58:52","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1918131,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDetermination of target genes associated with MASLD DAVs.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A-B)\u003c/strong\u003e Protein-protein interaction (PPI) networks of TFs significantly disrupted by DAVs in HepG2 (A) and LX-2 (B) cells, based on STRING database analysis. Edge thickness reflects confidence scores of interactions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C-D)\u003c/strong\u003e Wikipathways enrichment analysis of TFs significantly perturbed by DAVs in HepG2 (C) and LX-2 (D) cells.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(E)\u003c/strong\u003e Schematic diagram illustrating four approaches for assigning target genes to DAVs, including proximal genes, genes identified through Hi-C chromatin interactions, co-accessible genes, and eQTL-associated genes.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(F) \u003c/strong\u003eComparison of target gene numbers identified by the four approaches in various DAV sets. \u003cstrong\u003e(G) \u003c/strong\u003eDistribution of the number of target genes assigned to each DAV through four integrated methodologies.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(H) \u003c/strong\u003ePie chart illustrating the number of DAVs overlapping with liver eQTLs that exhibit concordant or discordant allele-specific MPRA activity relative to eQTL effect sizes.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(I)\u003c/strong\u003eVenn diagram showing the overlap of target genes assigned by different DAV subsets.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(J) \u003c/strong\u003eEnriched biological pathways associated with target genes from multiple DAV groups.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(K) \u003c/strong\u003emRNA expression of nominated target genes in HepG2 cell following introduction of CRISPRi machinery (n = 6). Data are shown as mean ± SEM (Student’s t test).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(L) \u003c/strong\u003eFold change in expression of nominated target genes for candidate sgRNAs identified by single-cell CROP-seq. (Significant, q \u0026lt; 0.1, red; Suggestive, \u003cem\u003ep \u003c/em\u003e\u0026lt; 0.05, blue).\u003c/p\u003e","description":"","filename":"FiguresandsFigures5.png","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/475fa7fd64ca466f9609e35e.png"},{"id":94161634,"identity":"72a868f7-8fe8-4fef-97a7-835c6eb77e78","added_by":"auto","created_at":"2025-10-23 04:58:51","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":2030808,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFunctional characterization of DAVs modulating APOA5, ANGPTL3, and LPL in triglyceride metabolism.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e Genomic visualization of rs1748197 and rs10889356 (highlighted in yellow) at the \u003cem\u003eANGPTL3\u003c/em\u003e locus, with annotations including chromatin accessibility, regulatory elements, and DAV signals.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B-C) \u003c/strong\u003eExpression levels of \u003cem\u003eANGPTL3\u003c/em\u003e in human liver samples across different genotypes of rs10889356 and rs1748197 from GTEX database.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(D-F) \u003c/strong\u003eRelative mRNA levels of \u003cem\u003eANGPTL3\u003c/em\u003ein HepG2 cells following CRISPR-mediated deletion (Del) (D), CRISPR interference (CRISPRi, CRi) knockdown (E), or CRISPR activation (CRISPRa, CRa) (F). Data represent mean ± SEM from 4-6 biological replicates.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(G) \u003c/strong\u003eFlow cytometry analysis of lipid accumulation in rs1748197-deleted HepG2 cells treated with or without PAOA for 24 h (n = 3).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(H) \u003c/strong\u003eRepresentative images of BODIPY staining in rs1748197-deleted HepG2 treated with PAOA for 24 h. Scale bar: 10 μm.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(I) \u003c/strong\u003eGenomic view of rs10790167 (yellow highlight) at the \u003cem\u003eAPOA5\u003c/em\u003e locus with relevant regulatory annotations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(J) \u003c/strong\u003emRNA expression of \u003cem\u003eAPOA5 \u003c/em\u003ein rs10790167-deleted HepG2. Data represent mean ± SEM from four biological replicates.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(K) \u003c/strong\u003eFlow cytometry analysis of lipid content in rs10790167-deleted HepG2 cells following 24 h PAOA treatment (n = 3).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(L) \u003c/strong\u003eRepresentative images of BODIPY staining in rs10790167-deleted HepG2 treated with PAOA for 24h. Scale bar: 10 μm.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(M) \u003c/strong\u003eMPRA-based regulatory activity of the G and A alleles of rs1441754 in HepG2 and LX-2 cell lines, showing allele-specific enhancer function.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(N) \u003c/strong\u003eGenomic visualization of rs1441754 (yellow box) at the \u003cem\u003eLPL\u003c/em\u003e locus, annotated with regulatory features and DAV assignment.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(O) \u003c/strong\u003emRNA expression of \u003cem\u003eLPL \u003c/em\u003eand\u003cem\u003e \u003c/em\u003efibrosis associated genes, (\u003cem\u003eTIMP1\u003c/em\u003e,\u003cem\u003e COL1A1\u003c/em\u003e and \u003cem\u003eACTA2\u003c/em\u003e) in rs1441754-deleted LX-2. Data represent mean ± SEM from three biological replicates.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(P) \u003c/strong\u003eRepresentative immunofluorescence images of COL1A1 and ACTA2 in rs1441754-deleted LX-2 cells after 48-hour TGFβ stimulation. Scale bar: 25 μm.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(Q) \u003c/strong\u003eSchematic summary model illustrating how DAVs at the\u003cem\u003e ANGPTL3\u003c/em\u003e, \u003cem\u003eAPOA5\u003c/em\u003e, and \u003cem\u003eLPL\u003c/em\u003e loci coordinate to regulate triglyceride metabolism and fibrogenic responses in hepatocytes and hepatic stellate cells.\u003c/p\u003e\n\u003cp\u003eData are shown as mean ± SEM (Student’s t test).\u003c/p\u003e","description":"","filename":"FiguresandsFigures6.png","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/702f830795e3993dd5e06ab6.png"},{"id":94162121,"identity":"dacdcb4d-1902-4078-874d-3c96cb84124d","added_by":"auto","created_at":"2025-10-23 05:06:52","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":1831042,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDAV rs474513 controls \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eSLC22A3\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e gene expression and lipid metabolism.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e Genomic visualization of rs474513 (highlighted in yellow) located within a hepatocyte-specific accessible chromatin peak at the \u003cem\u003eSLC22A3\u003c/em\u003e locus, with annotations including chromatin accessibility, regulatory elements, and DAV signals.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B) \u003c/strong\u003eExpression levels of \u003cem\u003eSLC22A3\u003c/em\u003eand \u003cem\u003eLPA\u003c/em\u003e in human liver samples across different genotypes of rs474513 from GTEX database.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C) \u003c/strong\u003ePredicted disruption of RUNX transcription factor binding by rs474513, based on TF motif analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(D)\u003c/strong\u003e Luciferase reporter assay showing allele-specific regulatory activity of rs474513 in 293T cells transfected with control, RUNX1, or RUNX2 expression plasmids (n = 3).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(E) \u003c/strong\u003emRNA expression of \u003cem\u003eSLC22A3\u003c/em\u003e and \u003cem\u003eLPA\u003c/em\u003e in HepG2 cell following deletion of a small genomic region containing rs474513 (n = 4).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(F) \u003c/strong\u003emRNA expression of \u003cem\u003eSLC22A3\u003c/em\u003e in WT and G\u0026gt;A heterozygous HepG2 cell (n = 3).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(G-H) \u003c/strong\u003eRepresentative BODIPY staining images showing lipid droplet accumulation in rs474513-deleted (G) and rs474513 G\u0026gt;A gene-edited (H) HepG2 cells following 24-hour treatment with PAOA. Scale bars, 10 µm.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(I-J) \u003c/strong\u003eFlow cytometry analysis of lipid content in HepG2 cells with rs474513 deletion (I) or G\u0026gt;A editing (J), with or without PAOA treatment for 24 h (n = 4).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(K) \u003c/strong\u003eEffect of RUNX inhibition on SLC22A3-mediated lipid accumulation in HepG2 cells. Representative BODIPY staining images following treatment with RO5-3335 (50 μM) or CADD522 (50 μM). Scale bars, 10 µm.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(L) \u003c/strong\u003eExpression of marker genes for various liver cell types in HLOs derived from hiPSCs treated with or without PAOA for 4 days (n = 3).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(M)\u003c/strong\u003e Representative images of BODIPY staining and immunofluorescence in HLOs. Top panel: BODIPY staining (green) showing lipid accumulation. Bottom panel: ALBUMIN (green) and ACTA2 (red) immunostaining. Scale bars, 10 µm.\u003c/p\u003e\n\u003cp\u003e(\u003cstrong\u003eN\u003c/strong\u003e) Odds ratios of polygenic risk scores (PRS) constructed from 114 LD-independent DAVs and matched MPRA control variants for MASLD-related traits in the Million Veteran Program (MVP). Variant weights were derived from effect sizes reported in the GWAS by Miao et al. as shown in Table S7. PRS were evaluated for four phenotypes: (1) MASLD-ALT, defined as ALT \u0026gt; 40 U/L in men or \u0026gt; 30 U/L in women; (2) MASLD-ICD, based on diagnostic ICD codes; (3) continuous ALT levels; and (4) fibrosis progression, assessed by the FIB-4 index. PRS were calculated in a cohort of 90,408 chronic ALT elevation (cALT) cases and 128,187 controls. Odds ratios represent the effect size per standard deviation increase in PRS.\u003c/p\u003e\n\u003cp\u003e(\u003cstrong\u003eO\u003c/strong\u003e) Odds ratios for each decile of PRSs constructed from 9 missense SNPs, DAVs, and MPRA-Ctrl variants, compared with their respective bottom deciles. For the top 4 deciles, the DAVs PRS demonstrated prediction power similar to that of the PRS constructed from 9 missense SNPs, with both outperforming the PRS constructed from MPRA-Ctrl in predicting MASLD-ALT.\u003c/p\u003e\n\u003cp\u003eData are presented as mean ± SEM, log odds ratios (log ORs), or odds ratios (ORs) with 95% confidence intervals (CIs), as indicated. Statistical significance was assessed using two-tailed Student’s t-test.\u003c/p\u003e","description":"","filename":"FiguresandsFigures7.png","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/f427b73941a7af5905b5847d.png"},{"id":103507250,"identity":"cf12d94b-c87f-4400-9b14-36560b9d1153","added_by":"auto","created_at":"2026-02-26 13:40:47","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":14729454,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/26b357d6-9918-424a-ae15-f739963563bd.pdf"},{"id":94161625,"identity":"2ec527b1-3e6c-4756-a90e-a223b0ee2c75","added_by":"auto","created_at":"2025-10-23 04:58:51","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":15656,"visible":true,"origin":"","legend":"Supplementary Table 1","description":"","filename":"TableS1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/05c95998495329ab00c3344b.xlsx"},{"id":94162113,"identity":"d54e2fe7-c0ad-467f-bcec-c6b46b607cea","added_by":"auto","created_at":"2025-10-23 05:06:51","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":39222,"visible":true,"origin":"","legend":"Supplementary Table 2","description":"","filename":"TableS2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/28ed16c000678dd7abbf266a.xlsx"},{"id":94161629,"identity":"f63db0ce-ed99-486f-aedf-dc9dbc3c6c64","added_by":"auto","created_at":"2025-10-23 04:58:51","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":46645,"visible":true,"origin":"","legend":"Supplementary Table 3","description":"","filename":"TableS3.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/cdb18a66f49f0fa0e4980ef1.xlsx"},{"id":94162118,"identity":"cfc498f9-34b1-4d58-9453-851357295196","added_by":"auto","created_at":"2025-10-23 05:06:51","extension":"xlsx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":37153,"visible":true,"origin":"","legend":"Supplementary Table 4","description":"","filename":"TableS4.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/480c242b324269e5b55c5bbc.xlsx"},{"id":94162122,"identity":"eeda9b79-a603-4485-b566-17885bf08e80","added_by":"auto","created_at":"2025-10-23 05:06:52","extension":"xlsx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":58929,"visible":true,"origin":"","legend":"Supplementary Table 5","description":"","filename":"TableS5.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/253bcc974497f7bd2049a765.xlsx"},{"id":94161645,"identity":"06b1732a-1893-4446-b568-e3e7d2fb609d","added_by":"auto","created_at":"2025-10-23 04:58:51","extension":"xlsx","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":15053,"visible":true,"origin":"","legend":"Supplementary Table 6","description":"","filename":"TableS6.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/bfc08cbd3cf616067ccf819a.xlsx"},{"id":94161646,"identity":"c4aa1658-951a-4d58-b7b9-01b8c74b2c01","added_by":"auto","created_at":"2025-10-23 04:58:51","extension":"xlsx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":24300,"visible":true,"origin":"","legend":"Supplementary Table 7","description":"","filename":"TableS7.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/0e6f1ca9a88f1a8c74d6d346.xlsx"},{"id":94161635,"identity":"5e52059c-1abd-468c-90e6-581f5b1fff52","added_by":"auto","created_at":"2025-10-23 04:58:51","extension":"xlsx","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":14804,"visible":true,"origin":"","legend":"Supplementary Table 8","description":"","filename":"TableS8.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/1c9a19d6407d26f71bfc6201.xlsx"},{"id":94161647,"identity":"03ca1baf-7ca0-4e7a-85ba-66e6c3555632","added_by":"auto","created_at":"2025-10-23 04:58:52","extension":"xlsx","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":31849,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Table 9\u003c/p\u003e","description":"","filename":"TableS9.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/5c8480de961c344fa701acf9.xlsx"},{"id":94161654,"identity":"b8e710a1-f3c8-499f-b602-eb9f3b0614a6","added_by":"auto","created_at":"2025-10-23 04:58:52","extension":"docx","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":150933,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Information\u003c/p\u003e","description":"","filename":"Supplementaryinformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/eea30d713852ebb1b2432220.docx"},{"id":94161649,"identity":"11cb2871-872f-40ee-aaa4-fd308eb991a2","added_by":"auto","created_at":"2025-10-23 04:58:52","extension":"pdf","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":7520166,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eExtended Data Figure Legends\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 1. Human liver cell types display unique patterns of chromatin accessibility.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A-F)\u003c/strong\u003e Quality control metrics for snATAC-seq data. Shown are the number of captured nuclei per sample (A), read distribution across genomic regions (B), transcription start site (TSS) enrichment scores (C), nucleosome signal profiles (D), percentage of reads mapping to called peaks (E), and relative cell type proportions across samples (F).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(G) \u003c/strong\u003eGenomic annotation of accessible peaks assigned to each major liver cell type.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(H)\u003c/strong\u003e UMAP visualization of fully integrated snATAC-seq data from normal and MASLD liver samples.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(I) \u003c/strong\u003eUMAP embedding visualized with coloration distinguishing between normal and MASLD individuals.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(J)\u003c/strong\u003e Comparison of cell type distributions between normal and MASLD liver tissues.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(K)\u003c/strong\u003e Genome browser tracks showing aggregated chromatin accessibility at marker gene loci across major liver cell types.\u003c/p\u003e\n\u003cp\u003eData are presented as mean ± SEM. Statistical significance was assessed using Student’s t-test.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 2. Human liver cell types exhibit distinct transcriptional regulatory program.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e Heatmap of cell type–specific chromatin accessibility peaks. Colors represent normalized marker peak intensities across major liver cell types.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B-C)\u003c/strong\u003e \u0026nbsp;Heatmap (B) and UMAP visualization (C) of enriched transcription factor (TF) motifs across each cell type. Colors indicate normalized motif enrichment scores. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(D-E)\u003c/strong\u003e Integrated analysis of our snATAC-seq data with published human liver snATAC-seq profiles from Kim et al. UMAP projection of merged cell types across both datasets (D). Correlation matrix comparing cell type annotations between the two studies (E).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(F)\u003c/strong\u003e Heatmap of cell type-specific marker gene activity derived from snATAC-seq gene scores. Colors denote normalized accessibility-derived gene scores.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(G)\u003c/strong\u003e UMAP plot displaying cell type-specific marker gene scores for\u003cem\u003e KRT7\u003c/em\u003e (cholangiocytes), \u003cem\u003eFLT1\u003c/em\u003e (endothelial cells), \u003cem\u003eALB\u003c/em\u003e (hepatocytes), \u003cem\u003eCD163\u003c/em\u003e (Kupffer cells), \u003cem\u003eCD8A\u003c/em\u003e (natural killer cells), and \u003cem\u003eNGF\u003c/em\u003e (stellate cells).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(H)\u003c/strong\u003e Heatmap illustrating alignment between cell type annotations from snATAC-seq and snRNA-seq datasets from Xiao et al.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 3. Enrichment of MASLD-associated variants in hepatocyte and stellate cells.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A-B)\u003c/strong\u003e Enrichment of GWAS variants associated with plasma triglycerides, body mass index (BMI), and MASLD in cell type-specific accessible chromatin regions. Regression coefficients (A) and enrichment scores (B) were calculated using linkage disequilibrium score regression (LDSC).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C-D)\u003c/strong\u003e Gene ontology (GO) enrichment analysis of MASLD-specific and normal-specific stellate cell subclusters identified in Figure 1F.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(E-F)\u003c/strong\u003e Pseudotime trajectory analysis of stellate cells inferred by StaVia (E), and correlation of StaVia pseudotime values with those derived from Monocle3 (F), demonstrating consistency across trajectory inference methods.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(G-H)\u003c/strong\u003e GO enrichment analysis of MASLD-specific and normal-specific hepatocyte subclusters shown in Figure 1L.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(I-J)\u003c/strong\u003e Pseudotime analysis of hepatocyte populations using StaVia (I) and comparison with Monocle3-based pseudotime estimates (J), supporting robustness of inferred dynamic transitions.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 4. Enrichment analysis of liver disease-associated variants within relevant liver cell types.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A-B)\u003c/strong\u003e UMAP projection of three transcriptionally distinct Kupffer cell subclusters (A) and their relative proportions across normal and MASLD liver tissues (B).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C-D)\u003c/strong\u003e Gene ontology (GO) enrichment analysis of normal-specific and MASLD-specific Kupffer cell clusters.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(E-F)\u003c/strong\u003e Trait Relevance Score (TRS) analysis of Kupffer cell subclusters for MASLD GWAS traits. UMAP visualization of SCAVENGE-inferred TRSs (E). Box plots showing TRS distributions across the three subclusters (F). Boxes represent the interquartile range (IQR), horizontal lines indicate the median, and whiskers denote 1.5× IQR.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(G)\u003c/strong\u003e Correlation between motif enrichment scores and MASLD GWAS variant enrichment within hepatocytes. Top 5 motifs were shown.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(H)\u003c/strong\u003e UMAP visualization of SCAVENGE TRSs for the cirrhosis trait across all liver cell types. \u003cstrong\u003e(I)\u003c/strong\u003e Correlation between TF motif enrichment and cirrhosis GWAS variant enrichment within hepatocytes. Top five enriched motifs are shown.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(J) \u003c/strong\u003eHepatocytes-specific gene regulatory networks (GRN) in Normal and MASLD condition. Each node in the network represents a gene or transcription factor, with its color and size denoting the expression level and PageRank centrality of the respective gene or transcription factor. The edges between nodes indicate the statistical significance of interactions between transcription factors and genes. Only the nodes identified as transcription factors are labeled in the network.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(K) \u003c/strong\u003eThe expression levels and motif activities (represented as chromVAR deviation Z-scores) of transcription factors between the GRNs of MASLD and Normal hepatocytes in extended data figure 4J.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 5. Design of the MPRA and assessment of its reproducibility.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e Schematic diagram outlining the strategy for selecting MASLD-associated genetic variants for inclusion in the MPRA library.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B)\u003c/strong\u003e Intersection of total MPRA variants with liver cell type-specific promoter and enhancer regions.\u003cstrong\u003e \u003c/strong\u003eCell type-specific promoter regions defined herein refer to promoter regions (2 kb upstream and 0.5 kb downstream of the TSS) that overlap with cell type-specific peaks, while the remaining cell type-specific peaks are designated as cell type-specific enhancers.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C-D)\u003c/strong\u003e Odds ratios (C) and (D) significance levels demonstrating the enrichment of total MPRA variants in liver cell type-specific promoter and enhancer regions versus random genomic regions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(E)\u003c/strong\u003e Histogram depicting the complexity of the MPRA library, featuring a median of 714 barcodes assigned per element.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(F)\u003c/strong\u003e GFP fluorescence intensities demonstrate higher enhancer activity for positive CRS, moderate activity for MPRA variants, and limited activity for negative CRS in HepG2 cells.\u003cstrong\u003e (G)\u003c/strong\u003e Quantitative MPRA activity across variant sets, including positive controls, negative controls, and both alleles of tested MPRA variants, measured in HepG2 cells.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(H)\u003c/strong\u003e Reproducibility analysis of MPRA performed in HepG2 cells across four biological replicates, showing high correlation and technical robustness.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 6. Characterization of DAVs identified by MPRA in HepG2.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e TF motif enrichment analysis of active genomic elements (at least one active allele) in HepG2 cells, using other inactive element as background.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B)\u003c/strong\u003e Manhattan plot of 5,369 MASLD variants assessed by MPRA. Red dots indicate genome-wide significant variants (\u003cem\u003ep\u003c/em\u003e \u0026lt; 1 × 10⁻\u003csup\u003e6\u003c/sup\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C)\u003c/strong\u003e Genomic distribution of all tested MPRA variants across chromosomes. DAVs (FDR \u0026lt; 0.01) are highlighted in red.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(D)\u003c/strong\u003e Pie chart showing the distribution of DAVs across different genomic categories (e.g., promoter, intronic, intergenic, enhancer regions).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(E)\u003c/strong\u003e Pie chart illustrating the number of functional genomic annotations associated with each DAV, including ATAC peaks from snATAC-seq data, liver eQTLs, chromatin interaction loci, disrupted TF binding motifs, and ChromHMM states.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(F-G)\u003c/strong\u003e Overlap of DAVs with liver cell type-specific chromatin accessibility peaks (F) and total peaks across all liver cell types (G).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(H)\u003c/strong\u003e Enrichment analysis showing the overrepresentation of DAVs in cell type-specific accessible chromatin regions relative to all tested MPRA variants. Enrichment is represented by -log10(\u003cem\u003ep\u003c/em\u003e) and visualized by color scale.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(I)\u003c/strong\u003e The enrichment of transcription factor motifs affected by DAVs, as identified by the motifbreakR algorithm.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 7. Visualizations of DAVs identified by MPRA in HepG2.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e Genomic browser view of the \u003cem\u003eGLDC\u003c/em\u003e locus. DAVs (highlighted in yellow) are shown within hepatocyte-specific accessible chromatin peaks, annotated enhancer regions, and chromatin loop anchors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B)\u003c/strong\u003e Genomic browser view of the \u003cem\u003eMLXIP\u003c/em\u003e locus. DAVs (highlighted in yellow) are located within hepatocyte-specific accessible chromatin peaks, annotated enhancer regions, and chromatin loop anchors.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 8. Motif activities of transcription factors influenced by DAVs in normal versus MASLD livers.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e UMAP visualization of DAVs-affected TF motifs across each cell type, including GABPA, HNF4G, KLF5, MYC, SP1, SP2, SREBF1, and SREBF2. Colors indicate normalized motif enrichment scores.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B) \u003c/strong\u003eBox plots showing the motif activity of DAVs-affected TFs in normal versus MASLD livers. Boxplots show median as a horizontal line, interquartile range as a box.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C-E)\u003c/strong\u003e UMAP visualization of DAVs-affected TF motifs within hepatocytes, including GABPA (C), CREB3 (D), and KLF15 (E). Colors indicate normalized motif enrichment scores.\u003cstrong\u003e \u003c/strong\u003eBox plots showing the motif activity of these DAVs-affected TFs in normal versus MASLD hepatocytes. Boxplots show median as a horizontal line, interquartile range as a box.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 9. Identification and characterization of cell type-specific DAVs.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e Volcano plot illustrating the activity of cell type-specific elements. Elements with significantly differential activity (BH corrected p value \u0026lt; 0.05) are highlighted in red and blue dots.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B) \u003c/strong\u003eTF\u003cstrong\u003e \u003c/strong\u003emotif enrichment analysis of LX-2-specific regulatory elements, indicating enrichment for fibrosis-associated TFs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C) \u003c/strong\u003eEnhancer activity for each element was plotted according to its activity rank in LX-2. Alleles were classified as Enhancers (red) if they had an adjusted \u003cem\u003ep \u003c/em\u003evalue \u0026lt; 0.05 and a log2 fold change \u0026gt; 0.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(D)\u003c/strong\u003e Intersection sizes of LX-2 DAVs overlapping with functional annotations such as ATAC\u0026nbsp; \u0026nbsp;peaks from snATAC-seq data, liver eQTLs, liver chromatin interactions, transcription factor binding motifs disruption, and ChromHMM states.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(E) \u003c/strong\u003ePie chart showing the number of functional annotations associated with each LX-2 DAV. \u003cstrong\u003e(F) \u003c/strong\u003eTF motif enrichment analysis of cell type-specific DAVs. Colors indicate significantly enriched motifs (q \u0026lt; 0.1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 10. Identification and characterization of stimuli-induced DAVs.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e Volcano plot illustrating the activity of PAOA-induced elements. Elements with significantly differential activity (BH corrected p value \u0026lt; 0.05) are highlighted in red and blue dots.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B-C) \u003c/strong\u003eTF\u003cstrong\u003e \u003c/strong\u003emotif enrichment analysis of PAOA-responsive regulatory elements. Red and blue colors denote significantly enriched motifs (q \u0026lt; 0.1). The FOXF1, BMAL1, HNF6 motifs highlighted in light blue and THRβ motif highlighted in light Salmon color indicate nominal significance (p \u0026lt; 0.01).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(D) \u003c/strong\u003eEnhancer activity for each element was plotted according to its activity rank in PAOA-treated HepG2. Alleles were classified as Enhancers (red) if they had an adjusted \u003cem\u003ep\u003c/em\u003e-value \u0026lt; 0.05 and a log2 fold change \u0026gt; 0.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(E) \u003c/strong\u003eThe enrichment of transcription factor motifs affected by PAOA-HepG2 DAVs, as identified by the SNP2TFBS algorithm.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(F)\u003c/strong\u003e Scatter plots illustrating the allele-differential activities of MPRA variants between PAOA- and Ctrl-treated conditions\u003cstrong\u003e.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(G)\u003c/strong\u003e Enhancer activity for each element was plotted according to its activity rank in TGFb-treated LX-2 cell. Alleles were classified as Enhancers (red) if they had an adjusted \u003cem\u003ep\u003c/em\u003e-value \u0026lt; 0.05 and a log2 fold change \u0026gt; 0.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(H)\u003c/strong\u003e Intersection sizes of TGFb-treated LX-2 DAVs overlapping with functional annotations such as ATAC peaks from snATAC-seq data, liver eQTLs, liver chromatin interactions, transcription factor binding motifs disruption, and ChromHMM states.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(I) \u003c/strong\u003eThe enrichment of transcription factor motifs affected by TGFb-treated LX-2 DAVs, as identified by the SNP2TFBS algorithm.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 11. DAVs-controlled transcriptional network.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e Protein-protein interaction network of all DAVs disrupted TFs from four MPRA datasets.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B)\u003c/strong\u003e WikiPathways enrichment of all DAVs disrupted TFs from four MPRA datasets.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 12. Determination of target genes associated with MASLD DAVs.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A)\u003c/strong\u003e Venn diagram illustrating the overlap of DAVs across different conditions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B) \u003c/strong\u003eComparison of target gene numbers identified by the four approaches.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C-F) \u003c/strong\u003eDistribution of the number of target genes assigned to each DAV in HepG2 (C), PAOA-treated HepG2 (D), LX-2 (E), and TGFb-treated LX-2 (F).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(G-J) \u003c/strong\u003ePie chart illustrating the number of DAVs overlapping with liver eQTLs that exhibit concordant or discordant allele-specific MPRA activity relative to eQTL effect sizes in HepG2 (G), PAOA-treated HepG2 (H), LX-2 (I), and TGFb-treated LX-2 (J).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(K) \u003c/strong\u003eGRN of DAVs target genes. Each node in the network represents a gene or transcription factor, with its color and size denoting the log2 fold change of expression level between MASLD and Normal, and PageRank centrality of the respective gene or transcription factor. The edges between nodes indicate the statistical significance of interactions between transcription factors and genes. Only the nodes identified as transcription factors are labeled in the network.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 13. CROP-seq of candidate DAVs.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(\u003cstrong\u003eA\u003c/strong\u003e) Immunofluorescence images displaying Cas9 expression in dCas9-KRAB/MeCP2 and dCas9-VPR knockin HepG2 cell line.\u003cstrong\u003e \u003c/strong\u003eScale bar: 200μm.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B) \u003c/strong\u003emRNA expression of nominated target genes in HepG2 cell following introduction of CRISPRa machinery (n = 6).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C)\u003c/strong\u003e The workflow outlining the selection criteria of CROP-seq experiment.\u003c/p\u003e\n\u003cp\u003e(\u003cstrong\u003eD\u003c/strong\u003e) Distribution of the number of cells targeted by each individual sgRNA in the CROP-seq screen.\u003c/p\u003e\n\u003cp\u003e(\u003cstrong\u003eE\u003c/strong\u003e) Total number of cells associated with each candidate DAV-targeting sgRNA across the dataset.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(F) \u003c/strong\u003eUMAP visualization of single-cell transcriptomic profiles from the CROP-seq experiment in HepG2 cells infected with lentiviral sgRNA constructs targeting DAV-associated regulatory regions.\u003c/p\u003e\n\u003cp\u003eData are expressed as mean ± SEM (Student’s t test).\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 14. Functional detection of DAVs in HepG2 cell line and LX-2 cell lines.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A) \u003c/strong\u003eAgarose gel electrophoresis of PCR-amplified genomic regions in stable DAV-deleted HepG2 or LX-2 cell lines. Genomic loci corresponding to rs1748197, rs10889356, rs10790167, rs1441754, and rs66477683 were used as PCR templates. Bands highlighted in black rectangles indicate successful deletion of the corresponding DAV region.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B)\u003c/strong\u003e Position weight matrix (PWM) of the MYC binding motif (minus strand) predicted to be disrupted by rs1748197.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C) \u003c/strong\u003emRNA expression of \u003cem\u003eANGPTL3\u003c/em\u003e and \u003cem\u003eUSP1\u003c/em\u003e in HepG2 cells following CRISPR-mediated deletion (Del), CRISPR interference (CRISPRi, CRi), or CRISPR activation (CRISPRa, CRa) (n = 3-6).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(D) \u003c/strong\u003ePWM of the GABPA binding motif (minus strand) predicted to be altered by rs10889356.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(E) \u003c/strong\u003eLuciferase reporter activity of the two alleles of rs10889356 in 293T cells expressing pLVX-Ctrl or pLVX-GABPA.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(F) \u003c/strong\u003eFlow cytometry analysis of lipid accumulation in rs10889356-deleted HepG2 cells treated with or without PAOA for 24 h (n = 3).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(G) \u003c/strong\u003eRepresentative images of BODIPY staining in rs10889356 deleted-HepG2 treated with PAOA for 24 h. Scale bars, 10 µm.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(H) \u003c/strong\u003eRepresentative images of BODIPY staining in 3D Liver Spheroids treated with PAOA for 48 h. Scale bars, 100 µm.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(I) \u003c/strong\u003ePWMs of RXR and PPARα binding motifs (minus strand) affected by rs10790167.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(J) \u003c/strong\u003eLuciferase reporter activity of rs1441754 alleles across three liver cell lines: LX-2, HepG2, and Huh7.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(K) \u003c/strong\u003eThe putative effect of rs1441754 on RUNX binding.\u003c/p\u003e\n\u003cp\u003eData are expressed as mean ± SEM (Student’s t test).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 15. Functional detection of rs66477683 in LX-2 cell line.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A) \u003c/strong\u003eGenomic browser view of the rs66477683 locus (highlighted in yellow) near the MERTK gene, annotated with regulatory features including chromatin accessibility, enhancer elements, and chromatin interactions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B) \u003c/strong\u003eLuciferase reporter activity of the two alleles of rs66477683 in 293T cells, assessing allele-specific regulatory function.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C) \u003c/strong\u003emRNA expression of \u003cem\u003eMERTK\u003c/em\u003e and fibrosis-related genes, including \u003cem\u003eTIMP1\u003c/em\u003e, \u003cem\u003eCOL1A1\u003c/em\u003e and \u003cem\u003eACTA2\u003c/em\u003e in rs66477683-deleted LX-2 (n = 3).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(D) \u003c/strong\u003eRepresentative immunofluorescence images of COL1A1 and ACTA2 in LX-2 cells treated with or without 20 ng/mL TGFβ for 48 h. Scale bars, 25 µm.\u003c/p\u003e\n\u003cp\u003eData are expressed as mean ± SEM (Student’s t test).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExtended Data Fig. 16. Functional detection of rs474513 in HepG2, Huh7 cell lines and iPSCs-derived hepatic-like cells.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(A) \u003c/strong\u003eAgarose gel electrophoresis of PCR-amplified genomic DNA from rs474513-deleted HepG2 cells. Bands highlighted in black rectangles indicate successful deletion of the DAV-containing region.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(B) \u003c/strong\u003emRNA expression of \u003cem\u003eSLC22A3\u003c/em\u003e in rs474513-deleted HepG2 cells treated with the RUNX2 inhibitor CADD522 (n = 4), as quantified by qRT-PCR and normalized to HPRT.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(C) \u003c/strong\u003emRNA expression of \u003cem\u003eSLC22A3\u003c/em\u003e and\u003cem\u003e LPA\u003c/em\u003e in HepG2 cells following CRISPR interference (CRISPRi)-mediated repression (n = 3-6).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(D) \u003c/strong\u003eSanger sequencing chromatogram confirming correction of the G/G genotype to G/A at rs474513 in edited cells.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(E) \u003c/strong\u003e\u003cem\u003eSLC22A3\u003c/em\u003e mRNA expression in HepG2 cells transfected with SLC22A3 overexpression plasmid for 24 h, followed by treatment with RUNX inhibitors (RO5-3335 for RUNX1 and CADD522 for RUNX2) for another 24 h.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(F) \u003c/strong\u003eQuantification of BODIPY intensity per cell from Figure 7K using ImageJ (n = 3 fields).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(G) \u003c/strong\u003eAgarose gel electrophoresis of PCR products from rs474513-deleted iPSC lines. Bands within black rectangles denote successful deletion.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(H) \u003c/strong\u003eRelative mRNA expression of \u003cem\u003eSLC22A3\u003c/em\u003e in rs474513-deleted iPSC lines.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(I) \u003c/strong\u003eRelative mRNA expression of marker genes at different stages of hepatic-like cell differentiation derived from iPSCs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e(J)\u003c/strong\u003e Comparison of odds ratios for PRS constructed from 114 LD-independent DAVs for MASLD traits in the All of Us cohort (weighted by MVP GWAS, Vujkovic et al) versus PRS9, created from 9 Missense SNPs (Methods), and published PRS (PGS002732 and PGS000655) from the PGS Catalog.\u003c/p\u003e\n\u003cp\u003eData are expressed as mean ± SEM (Student’s t test).\u003c/p\u003e","description":"","filename":"ExtendFigures.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6984670/v1/69e62a08e9be4812ec5dd6ae.pdf"}],"financialInterests":"\u003cb\u003eYes\u003c/b\u003e there is potential Competing Interest.\nMAL is on a scientific advisory board of Pfizer and is co-founder and serves on the scientific advisory board of Flare Therapeutics. DJR is on the scientific advisory board for Alnylam, Novartis, and Verve. All other authors have no competing interests to declare.","formattedTitle":"Integrative functional genomics reveals transcriptional regulatory function of risk alleles for metabolic liver disease","fulltext":[{"header":"Introduction","content":"\u003cp\u003eMetabolic dysfunction-associated steatotic liver disease (MASLD), the most common chronic liver condition, encompasses a spectrum from metabolic dysfunction-associated steatotic liver (MASL), characterized by simple lipid accumulation, to advanced metabolic dysfunction-associated steatohepatitis (MASH), marked by progressive inflammation and fibrosis\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. Given the limited therapeutic options for MASLD, there is an urgent need to elucidate its underlying causal mechanisms. MASLD arises from the complexed interplay between environmental factors and genetic predisposition, with heritability estimates ranging from 20% to 70%\u003csup\u003e2\u003c/sup\u003e. Large-scale genome-wide association studies (GWAS) have identified nearly 100 loci associated with MASLD risk, but most variants lie in noncoding regions with unknown functions\u003csup\u003e\u003cspan additionalcitationids=\"CR4 CR5 CR6\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. Furthermore, high linkage disequilibrium (LD) and genetic pleiotropy complicate the identification of causal variants and target genes. Numerous computational methodologies, including fine-mapping, expression quantitative trait loci (eQTL) analysis, chromatin accessibility quantitative trait loci (caQTL) analysis, and deltaSVM algorithms, have been employed to identify causal functional variants, but they have demonstrated limited predictive power\u003csup\u003e\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eNoncoding genetic variants can influence gene expression by altering chromatin accessibility and transcription factor binding at gene regulatory elements (GREs), such as enhancers\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. Massively parallel reporter assays (MPRA) offer a scalable experimental approach to test the enhancer activity of risk alleles in disease-relevant cell types, organoids, or organisms under various environmental conditions\u003csup\u003e\u003cspan additionalcitationids=\"CR14\" citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. Since GREs are often cell-type-specific, it is crucial to understand the role of disease-associated variants within relevant cell types. While many studies using single-cell/nuclei RNA sequencing (scRNA-seq or snRNA-seq) have shown that MASLD pathophysiology involves multiple liver cell types\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e,\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e, the chromatin landscape heterogeneity in MASLD patient livers has been rarely explored. This understanding is essential for interpreting the molecular mechanisms underlying noncoding MASLD risk variants.\u003c/p\u003e\u003cp\u003eHere, we profiled the chromatin accessibility landscape of liver nuclei from MASLD patients and found that accessible regions were enriched for MASLD risk variants and bound by cell type-specific transcription factors. MPRA screening identified hundreds of differential activity variants (DAVs) with context-specific regulatory activity that perturb transcriptional networks associated with liver pathogenesis. Integration with liver eQTL data, chromatin interaction maps, and single-cell CRISPRi screening enabled the assignment of target genes to these regulatory variants. Notably, we validated key DAVs that modulate lipid metabolism and drive hepatic stellate cell activation, the hallmarks of MASLD pathology. Importantly, these DAVs exhibit predictive power in distinguishing MASLD disease risk. Thus, our multimodal analyses reveal how noncoding genetic variation drives MASLD progression through disruption of cell type- and state-specific regulatory programs.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eCell type-specific regulation of chromatin accessibility by MASLD risk variants\u003c/h2\u003e\u003cp\u003eTo delineate the regulatory program of MASLD-associated genetic noncoding variants, we profiled the chromatin landscape of MASLD patient liver. We performed single-nucleus ATAC sequencing (snATAC-seq) on liver nuclei from 12 individuals (6 Normal and 6 MASLD) \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA\u003cb\u003e)\u003c/b\u003e, and identified 69,595 high-quality nuclei and classified them into 15 distinct cell clusters \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB \u003cb\u003eand Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA-\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eK\u003cb\u003e)\u003c/b\u003e. Each cluster displayed a unique chromatin accessibility landscape and TF motif enrichment signature, reflecting robust cell identity-specific regulatory programs \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA-\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC\u003cb\u003e)\u003c/b\u003e. Our cell type annotations were cross-validated using a recently published human liver snATAC reference dataset \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD-\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eE\u003cb\u003e)\u003c/b\u003e\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. Furthermore, by inferring the gene scores of marker genes of each cluster \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eF\u003cb\u003e)\u003c/b\u003e, we identified clusters representing hepatocytes (\u003cem\u003eALB\u003c/em\u003e), cholangiocytes (\u003cem\u003eKRT7\u003c/em\u003e), Kupffer cells (\u003cem\u003eCD163\u003c/em\u003e), stellate cells (\u003cem\u003eNGF\u003c/em\u003e), endothelial cells (\u003cem\u003eFLT1\u003c/em\u003e), and T/NK/B cells (\u003cem\u003eCD8A\u003c/em\u003e), with each cell cluster expressing a unique transcriptional program \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB \u003cb\u003eand Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eG\u003cb\u003e)\u003c/b\u003e. We observed highly specific correlations between marker genes defined by snATAC-seq and previous snRNA-seq from human MASLD liver \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eH\u003cb\u003e)\u003c/b\u003e\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. As expected, MASLD liver samples consisted of higher percentages of cholangiocytes and stellate cells, and decreased hepatocyte and endothelial cells, compared to normal livers \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC \u003cb\u003eand Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eH-\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eJ\u003cb\u003e)\u003c/b\u003e\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eWe next sought to test whether MASLD-associated variants were enriched in liver regulatory elements. Stratified linkage disequilibrium score regression (LDSC) analysis demonstrated enrichment of MASLD GWAS variants in hepatocyte and stellate cell chromatin peaks, while plasma triglyceride trait-associated variants were enriched in accessible genomic regions of hepatocytes and Kupffer cells \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA-\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB\u003cb\u003e)\u003c/b\u003e. In contrast, no enrichment was observed for variants associated with body mass index (BMI) traits. To further investigate the heterogeneity of genetic association enrichment, we applied SCAVENGE, an algorithm we developed that integrates scRNA-seq and GWAS data to identify disease-associated genetic variants at single-cell resolution\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. Our analysis revealed that liver fat trait-associated variants were specifically enriched in hepatocyte peaks, whereas MASLD-associated variants displayed a broader pattern of enrichment across various cell types, including hepatocytes, stellate cells, and Kupffer cells \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eD-\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eE\u003cb\u003e)\u003c/b\u003e, suggesting the genetic involvement of non-parenchymal cells in addition to hepatocytes in MASLD.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eWe then focused on stellate cells, the primary effector cells in MASLD progression\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e, and identified three sub-clusters with varying chromatin accessibility. Based on the cluster composition of cell origin, we termed these three stellate cell clusters as S-MASLD, S-Normal, and S-Common \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eF-\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eG\u003cb\u003e)\u003c/b\u003e. The proportion of S-MASLD cells was higher in MASLD liver compared to normal liver. Genomic region enrichment analysis of S-MASLD-specific peaks revealed pathways related to extracellular matrix organization and inflammatory responses \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC-\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD\u003cb\u003e)\u003c/b\u003e. Strikingly, MASLD genetic risk variants were differentially enriched among stellate subtypes, with S-MASLD cells exhibiting the highest SCAVENGE trait relevance scores (TRS) \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eH\u003cb\u003e)\u003c/b\u003e. Pseudotime trajectory analysis, validated by two independent algorithms, revealed a continuous activation process across stellate cell states, during which MASLD risk variants became progressively enriched \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eI-\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eK \u003cb\u003eand Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eE-\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eF\u003cb\u003e)\u003c/b\u003e. Beyond stellate cells, we observed substantial chromatin heterogeneity among hepatocyte subpopulations. Disease-enriched hepatocyte states showed preferential localization of MASLD-associated risk variants \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eL-\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eO \u003cb\u003eand Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eG-\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eJ\u003cb\u003e)\u003c/b\u003e. A similar trend was observed in Kupffer cells, where disease-associated Kupffer cells exhibited notably higher SCAVENGE TRS scores \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA-\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eF\u003cb\u003e)\u003c/b\u003e. Collectively, these findings indicate that MASLD risk variants exert cell type- and cell state-specific effects on the hepatic regulatory landscape to drive disease progression.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eGiven our ability to simultaneously map GWAS variant enrichments using SCAVENGE and TF motif activities using chromVAR at single-cell resolution, we next performed integrative analyses to gain mechanistic insight into how noncoding risk variants might influence transcriptional regulation in MASLD\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. We observed a positive correlation between MASLD enrichment TRS scores and the motif activities of TFs related to metabolism and fibrosis, such as SNAI2, across all profiled nuclei \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eP\u003cb\u003e)\u003c/b\u003e\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. A similar trend was seen when the analysis was performed in hepatocytes alone, where CEBPA motif activities were positively correlated with MASLD trait TRS scores \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eG\u003cb\u003e)\u003c/b\u003e. Additionally, integrative analysis of cirrhosis, a more advanced stage of MASLD, revealed the strongest positive correlations with TFs like CEBPA, HNF4A, and NR2F6, and negative correlations with TFs such as GABPA \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eQ \u003cb\u003eand Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eH-\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eI\u003cb\u003e)\u003c/b\u003e, which is involved in oxidative phosphorylation\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. To further dissect the regulatory architecture of MASLD progression, we integrated snRNA-seq and snATAC-seq data to construct hepatocyte-specific transcriptional regulatory networks using Pando\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. We revealed extensive disease-dependent transcriptional regulatory network rewiring, characterized by altered expression and motif enrichment of key TFs implicated in MASLD pathogenesis, including FOXO3, MLXIPL, PROX1, PPAR, HNF1B/4A, and TEAD1 \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eJ-K\u003cb\u003e)\u003c/b\u003e.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eIdentify and characterize the candidate functional variants using MPRA\u003c/h3\u003e\n\u003cp\u003eTo systematically identify functional noncoding variants associated with MASLD, we designed an MPRA library that includes both alleles of 5,369 noncoding variants in strong linkage disequilibrium (r\u0026sup2; \u0026gt;0.8) with genome-wide significant as well as subthreshold significant variants from recent GWASs \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA \u003cb\u003eand Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e)\u003c/b\u003e\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan additionalcitationids=\"CR26 CR27 CR28 CR29\" citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e. Notably, we included subthreshold variants in the MPRA library based on their potential to modulate enhancer function and contribute to MASLD risk, despite not surpassing conventional significance thresholds for GWAS\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e,\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e. Interestingly, these MPRA candidate variants were predominantly enriched in regions of open chromatin within primary liver cell types, such as hepatocytes and immune cells \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB-\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD\u003cb\u003e)\u003c/b\u003e. The MPRA was conducted in disease-relevant hepatic cell lines, specifically the hepatocyte cell line HepG2 and the stellate cell line LX-2, both with and without environmental stimuli (fatty acid and TGFβ) to simulate environmental risk factors associated with MASLD \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA\u003cb\u003e)\u003c/b\u003e. The MPRA library was synthesized with a median library complexity of 714 barcodes per allele and included 40 positive regulatory elements predicted to be active in hepatocytes \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eE\u003cb\u003e)\u003c/b\u003e. The positive regulatory elements demonstrated elevated transcriptional activity, whereas the negative regulatory elements exhibited minimal activity in HepG2 cells, validating the robustness of the MPRA approach \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eF-\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eG\u003cb\u003e)\u003c/b\u003e. MPRA was performed with four replicates for each condition, and exhibited high reproducibility, as exemplified in HepG2 \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eH\u003cb\u003e)\u003c/b\u003e. Overall, 2430 regulatory elements were transcriptionally active for at least one allele in HepG2 and enriched for metabolic TFs such as HNF4, CEBPD and TRβ \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB \u003cb\u003eand Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA\u003cb\u003e)\u003c/b\u003e. We subsequently assessed allele-specific activity within these active elements, identifying 365 differential activity variants (DAVs) [FDR\u0026thinsp;\u0026lt;\u0026thinsp;0.01] distributed across 21 chromosomes \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC, \u003cb\u003eExtended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eB-\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eC \u003cb\u003eand Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e)\u003c/b\u003e. To validate the sensitivity and specificity of the DAVs identified by MPRA, we randomly selected 10 DAVs with high allele-specific activity and examined their activities individually. Notably, eight out of ten DAVs demonstrated the expected allele-specific activity in HepG2 cells \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD\u003cb\u003e)\u003c/b\u003e, confirming the reliability of the MPRA.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eWe next characterized the properties of the DAVs through integrative analysis, incorporating chromatin accessibility information identified by snATAC-seq, liver eQTL data, liver or HepG2 chromatin interaction data, TF motif disruption, and liver chromatin functional annotations\u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e,\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. Nearly 90% of DAVs were located within active genomic regions, and 297 out of 365 DAVs overlapped with two or more functional annotations \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eE \u003cb\u003eand Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eD-\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eE\u003cb\u003e)\u003c/b\u003e. Of particular interest were 87 DAVs that were located in chromatin-accessible regions across various cell types, with higher enrichment observed in open chromatin regions of hepatocytes, stellate cells, and cholangiocytes \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eF \u003cb\u003eand Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eF-\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eH\u003cb\u003e)\u003c/b\u003e. For example, several DAVs located in the \u003cem\u003eGLDC\u003c/em\u003e and \u003cem\u003eMLXIP\u003c/em\u003e loci reside within hepatocyte-specific chromatin peaks and were associated with strong enhancer signals \u003cb\u003e(Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eA-\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eB\u003cb\u003e)\u003c/b\u003e. Notably, 274 DAVs altered the motif activity of specific TFs. We observed a positive correlation between TF binding disruption scores and MPRA effect sizes for the DAVs \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eG\u003cb\u003e)\u003c/b\u003e. Specifically, HNF4G, SREBF1/2, and ETS (Erythroblast Transformation-Specific) family binding sites were enriched for DAVs, as determined using two different algorithms, SNP2TFBS and MotifbreakR \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eH \u003cb\u003eand Extended Data\u003c/b\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eI\u003cb\u003e)\u003c/b\u003e. These results are consistent with the established roles of HNF4 and SREBF1/2 in liver metabolism and underscore the involvement of ETS family transcription factors in MASLD progression\u003csup\u003e\u003cspan additionalcitationids=\"CR36 CR37\" citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e. Furthermore, we observed that motif activities of HNF4G and SREBF1/2 were higher in the MASLD liver compared to the normal liver, whereas the ETS family member GABPα binding activity was reduced in MASLD liver \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eI \u003cb\u003eand Extended Data Fig.\u0026nbsp;8A-8E)\u003c/b\u003e. Together, these functional annotations allow us to prioritize potential regulatory variants and nominate their target genes with greater precision.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\n\u003ch3\u003eDAVs exhibit cell type-specific regulatory function\u003c/h3\u003e\n\u003cp\u003eThe progression of MASLD involves multiple liver cell types, including the activation of hepatic stellate cells. We hypothesized that MASLD risk variants modulate the activity of regulatory elements in a cell type-specific manner. MPRA analysis revealed substantial differences in regulatory element activity between human hepatocyte cells line HepG2 and stellate cell line LX-2, in line with the cell type-specific chromatin accessibility observed \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA \u003cb\u003eand Extended Data Fig.\u0026nbsp;9A)\u003c/b\u003e. Regulatory elements specific to HepG2 cells were enriched for motifs associated with PPARα and ERRγ, key nuclear receptors involved in liver metabolism. In contrast, LX-2-specific elements contained binding sites for fibrosis-related TFs, such as STAT6 \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB \u003cb\u003eand Extended Data Fig.\u0026nbsp;9B)\u003c/b\u003e. These cell type-specific regulatory element activities resulted in more than half of the DAVs exhibiting allele-specific activity exclusively in either HepG2 or LX-2 cells \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC-\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eE \u003cb\u003eand Extended Data Fig.\u0026nbsp;9C)\u003c/b\u003e. In total, we identified 464 DAVs in LX-2 cells, with the majority being liver eQTLs, located in epigenomic regions, or disrupting specific TF binding motifs \u003cb\u003e(Extended Data Fig.\u0026nbsp;9D-9E and Table \u003cspan refid=\"MOESM3\" class=\"InternalRef\"\u003eS3\u003c/span\u003e)\u003c/b\u003e. Notably, LX-2-specific DAVs disrupted the binding sites of STAT, TEAD, and TCF21 \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eF \u003cb\u003eand Extended Data Fig.\u0026nbsp;9F)\u003c/b\u003e, all of which have been implicated in liver fibrosis\u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e. For instance, the genetic variant rs6726823 near the \u003cem\u003eCASP8\u003c/em\u003e gene exhibited significant allele-specific activity in LX-2 but showed far less differential activity in HepG2 \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eG\u003cb\u003e)\u003c/b\u003e. Interestingly, the predicted target gene of rs6726823, \u003cem\u003eCASP8\u003c/em\u003e, has been functionally implicated in hepatic stellate cell homeostasis and liver fibrosis in MASLD mouse models\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e,\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e. Mechanistically, we found that rs6726823 resides in stellate-specific chromatin peaks and may regulate \u003cem\u003eCASP8\u003c/em\u003e expression by affecting the binding motif of IRF, a transcription factor family implicated in liver fibrosis \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eH-\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eI\u003cb\u003e)\u003c/b\u003e\u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e. In contrast, the variant rs13150068 displayed allele-specific activity exclusively in HepG2 cells, with no notable differential activity in LX-2 \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eJ\u003cb\u003e)\u003c/b\u003e. This variant is located in a hepatocyte-specific open chromatin region near \u003cem\u003eHSD17B13\u003c/em\u003e, which is a liver-enriched gene involved in lipid droplet metabolism and strongly associated with MASLD pathogenesis \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eK\u003cb\u003e)\u003c/b\u003e\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e. Motif analysis suggests that rs13150068 may exert its regulatory effect by disrupting the binding site of SREBP2, a master transcriptional regulator of lipid biosynthesis \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eL\u003cb\u003e)\u003c/b\u003e. These findings underscore the importance of selecting disease-relevant cell lines to accurately identify bona fide functional variants.\u003c/p\u003e\n\u003ch3\u003eContext-dependent activity of DAVs\u003c/h3\u003e\n\u003cp\u003eSince the progression of MASLD is shaped by the complex interplay between genetic predisposition and environmental factors\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e, we further delineated the functional variants under various environmental stimuli, including oleic acid and palmitic acid (PAOA), and TGFβ to mimic the metabolic stress in MASLD livers\u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. Hundreds of regulatory elements were induced in HepG2 under PAOA condition, with PAOA-induced elements being enriched for the binding motifs of metabolic nuclear receptors, ERRα and TRβ \u003cb\u003e(Extended Data Fig.\u0026nbsp;10A-10C)\u003c/b\u003e. A total of 338 DAVs were identified in PAOA-treated hepatocytes, with most of them having at least two functional annotations \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA-\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB \u003cb\u003eand Extended Data Fig.\u0026nbsp;10D-10E and Table \u003cspan refid=\"MOESM4\" class=\"InternalRef\"\u003eS4\u003c/span\u003e)\u003c/b\u003e. Notably, one-third of the DAVs were stimulus-dependent, displaying variable allele-specific activity between control and PAOA conditions \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC \u003cb\u003eand Extended Data Fig.\u0026nbsp;10F)\u003c/b\u003e. Of note, these PAOA-specific DAVs appeared to preferentially affect the PPARE and RUNX1 motifs \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eD\u003cb\u003e)\u003c/b\u003e. In particular, rs12406250 exhibited higher allele-specific activity under PAOA conditions compared to control \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eE\u003cb\u003e)\u003c/b\u003e. This variant is located in an intronic open chromatin peak of the \u003cem\u003eEPHA2\u003c/em\u003e gene, whose expression and signaling are known to be altered during MASLD pathogenesis \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eF\u003cb\u003e)\u003c/b\u003e\u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e. Moreover, the intensity of the peak containing rs12406250 was elevated in MASLD-hepatocytes relative to normal hepatocytes \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eF\u003cb\u003e)\u003c/b\u003e, suggesting the role of rs12406250 in regulating the chromatin accessibility in MASLD.\u003c/p\u003e\u003cp\u003eWe further identified 614 DAVs in LX-2 cells in response to TGFβ treatment, and these DAVs likely exert their regulatory effects by modulating chromatin accessibility and TF binding dynamics \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eG \u003cb\u003eand Extended Data Fig.\u0026nbsp;10G-10I and Table \u003cspan refid=\"MOESM5\" class=\"InternalRef\"\u003eS5\u003c/span\u003e)\u003c/b\u003e. Notably, most of these DAVs exhibited differential allele-specific activity compared to the basal state, suggesting that TGFβ alters the genomic binding landscape and reveals context-specific variant function \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eH-\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eJ\u003cb\u003e)\u003c/b\u003e. For example, rs558702, a DAV located within a stellate cell-specific open chromatin region, exhibited enhancer activity exclusively upon TGFβ treatment, highlighting its context-specific regulatory role \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eK-\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eL\u003cb\u003e)\u003c/b\u003e. Together, these findings underscore how environmental cues such as PAOA and TGFβ exposure can unmask the functional consequences of noncoding variants and provide mechanistic insight into gene-environment interactions driving MASLD pathogenesis.\u003c/p\u003e\n\u003ch3\u003eDisease-associated transcriptional programs driven by DAVs\u003c/h3\u003e\n\u003cp\u003eTo elucidate the transcriptional networks regulated by context-specific DAVs, we systematically analyzed the TFs disrupted by DAVs in HepG2 and LX-2 cells. Protein-protein interaction network analysis revealed that TFs perturbed by HepG2-DAVs formed a tightly connected transcriptional module centered around SREBP signaling and lipid metabolism, whereas LX-2-perturbed TFs were enriched in a network linked to TGFβ signaling \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA-\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD \u003cb\u003eand Extended Data Fig.\u0026nbsp;11A-11B)\u003c/b\u003e. These results indicate that MASLD-associated DAVs converge on distinct, cell-type-specific transcriptional programs, with TFs functioning cooperatively within coordinated, disease-relevant regulatory circuits.\u003c/p\u003e\u003cp\u003eTo accurately identify the target genes of these DAVs, we integrated multiple orthogonal datasets, including public liver eQTL database, chromatin interaction data, and co-accessibility analyses from snATAC, as well as proximal gene information \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eE\u003cb\u003e)\u003c/b\u003e. This comprehensive approach enabled us to assign between 794 and 1,009 target genes for different set of DAVs, with 995 DAVs having at least one associated target gene \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eF-\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eG \u003cb\u003eand Extended Data Fig.\u0026nbsp;12A-12F)\u003c/b\u003e. Among these, 338 DAVs overlapped with known liver eQTLs, and notably, 64% exhibited concordant effect directions between reporter activity and eQTL effect size \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eH \u003cb\u003eand Extended Data Fig.\u0026nbsp;12G-12J)\u003c/b\u003e. The remaining 36% discordance likely reflects the cellular heterogeneity of liver tissue and the context-dependent nature of chromatin regulation. Interestingly, gene ontology analysis of these potential target genes revealed many biological processes associated with MASLD progression, including lipid metabolism, immune responses, and the TGFβ signaling pathway \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eI-\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eJ\u003cb\u003e)\u003c/b\u003e. Moreover, transcriptional regulatory network analysis revealed that many DAV target genes are regulated by key MASLD-related TFs, including SREBF1, MLXIPL (ChREBP), ERG, and TCF7L2 \u003cb\u003e(Extended Data Fig.\u0026nbsp;12K)\u003c/b\u003e.\u003c/p\u003e\u003cp\u003eTo validate the functional predictions, we deactivated the genomic regions spanning 8 randomly selected DAVs using CRIPSR interference (CRISPRi) approach and examined the expression levels of the nominated target genes. Over 90% of the potential target genes showed reduced expression levels, consistent with predictions \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eK \u003cb\u003eand Extended Data Fig.\u0026nbsp;13A)\u003c/b\u003e. Conversely, when CRIPSR activation (CRISPRa) machinery was introduced into the cells, most target genes exhibited increased expression levels \u003cb\u003e(Extended Data Fig.\u0026nbsp;13A-13B)\u003c/b\u003e. To systematically identify the target genes of MASLD-associated DAVs, we employed a CROP-seq approach that integrates scRNA-seq with CRISPR screening\u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e. A lentiviral-based CROP-seq library was generated with high complexity and uniformity, comprising 40 sgRNAs targeting 20 genomic regions spanning MASLD-associated DAVs with strong allelic effects and well-annotated metabolic target genes \u003cb\u003e(Extended Data Fig.\u0026nbsp;13C and Table \u003cspan refid=\"MOESM6\" class=\"InternalRef\"\u003eS6\u003c/span\u003e)\u003c/b\u003e. scRNA-seq of CROP library-infected HepG2 cells revealed 67,700 high-quality cells and identified 18 distinct cell clusters \u003cb\u003e(Extended Data Fig.\u0026nbsp;13D-13F)\u003c/b\u003e. We then associated the DAVs with the previously nominated target genes and identified 25 unique gene-sgRNA pairs, where the expression levels of the target genes were significantly lower in DAVs sgRNA-infected cells compared to all other sgRNA-infected cells \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eL\u003cb\u003e)\u003c/b\u003e. This analysis offers a robust strategy for narrowing down the pool of target genes associated with functional variants, providing mechanistic insights into MASLD progression.\u003c/p\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003eFunctional characterization of DAVs modulating APOA5, ANGPTL3, and LPL in triglyceride metabolism\u003c/h2\u003e\u003cp\u003eWith the identification of potential functional DAVs and their annotated target genes, we next investigated how these DAVs influence lipid metabolism, a key aspect of MASLD, in the liver. APOA5, ANGPTL3/8, and LPL form a tightly regulated network that plays a central role in triglyceride metabolism and systemic lipid homeostasis\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e. We then focused on two DAVs (rs1748197 and rs10889356) that are near the \u003cem\u003eANGPTL3\u003c/em\u003e locus \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA\u003cb\u003e)\u003c/b\u003e, a well-known inhibitor of LPL activity\u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e, and have been previously identified as risk variants associated with MASLD \u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. Although neutralizing antibodies targeting ANGPTL3 have been approved to reduce lipoprotein levels in individuals with homozygous familial hypercholesterolemia, the inhibition of hepatic ANGPTL3 using antisense oligonucleotide (ASO) drug Vupanorsen unexpectedly led to liver lipid accumulation, ultimately resulting in the termination of the clinical trial\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e,\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e. We discovered that DAV rs1748197 is located in the promoter region of the \u003cem\u003eANGPTL3\u003c/em\u003e gene, and confirmed that rs1748197 acts as an eQTL variant for \u003cem\u003eANGPTL3\u003c/em\u003e \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA-\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eB\u003cb\u003e)\u003c/b\u003e. Deletion of a small genomic region containing rs1748197 or deactivation of this genomic region using CRISPRi reduced the expression of \u003cem\u003eANGPTL3\u003c/em\u003e, while genomic activation using CRISPRa enhanced the expression of \u003cem\u003eANGPTL3\u003c/em\u003e \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eD-\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eF \u003cb\u003eand Extended Data Fig.\u0026nbsp;14A)\u003c/b\u003e. Mechanistically, rs1748197 is predicted to alter the binding motif of MYC, a transcription factor implicated in the pathogenesis of alcoholic liver disease \u003cb\u003e(Extended Data Fig.\u0026nbsp;14B)\u003c/b\u003e\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eMeanwhile, another DAV rs10889356 resides in an enhancer region that interacts with the \u003cem\u003eANGPTL3\u003c/em\u003e promoter \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA\u003cb\u003e)\u003c/b\u003e. Public liver eQTL data reveals a strong association between rs10889356 and \u003cem\u003eANGPTL3\u003c/em\u003e expression, with livers carrying G allele having higher \u003cem\u003eANGPTL3\u003c/em\u003e expression \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eC\u003cb\u003e)\u003c/b\u003e. Indeed, inhibition or activation of this enhancer region reduced or increased \u003cem\u003eANGPTL3\u003c/em\u003e expression, respectively \u003cb\u003e(Extended Data Fig.\u0026nbsp;14C)\u003c/b\u003e. Motif analysis further revealed that rs10889356 perturbs a GABPA binding motif, and reporter assays confirmed that the G allele promotes enhancer activity, likely through increased GABPA recruitment \u003cb\u003e(Extended Data Fig.\u0026nbsp;14D-14E)\u003c/b\u003e. Functionally, deletion of the small genomic regions containing rs1748197 and rs10889356 individually in HepG2 cells resulted in greater lipid accumulation compared to wild-type HepG2 cells under both basal and PAOA-treated conditions \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eG-\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eH \u003cb\u003eand Extended Data Fig.\u0026nbsp;14F-14G)\u003c/b\u003e. These findings were further confirmed in these DAVs-deleted hepatocyte spheroids treated with PAOA \u003cb\u003e(Extended Data Fig.\u0026nbsp;14H)\u003c/b\u003e. These data suggest that MASLD-associated DAVs rs1748197 and rs10889356 modulate the expression of \u003cem\u003eANGPTL3\u003c/em\u003e, which negatively regulates hepatic lipid metabolism.\u003c/p\u003e\u003cp\u003eWe further identified rs10790167 as a HepG2-specific DAV with robust allele-specific activity and no significant regulatory effect in LX-2 cells \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eI\u003cb\u003e)\u003c/b\u003e. This variant is located in a hepatocyte-specific open chromatin region and establishes an enhancer-promoter (E-P) interaction with the promoter of \u003cem\u003eAPOA5\u003c/em\u003e, which is a known activator of LPL activity and closely linked to MASLD progression\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e. Motif analysis revealed that rs10790167 disrupted the binding motifs of several nuclear receptors, including PPAR and RXR \u003cb\u003e(Extended Data Fig.\u0026nbsp;14I)\u003c/b\u003e. Deletion of rs10790167 significantly reduced the expression of \u003cem\u003eAPOA5\u003c/em\u003e and led to marked accumulation of lipid droplets in HepG2 cells, suggesting a causal role for this variant in dysregulated lipid metabolism \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eJ-\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eL\u003cb\u003e)\u003c/b\u003e.\u003c/p\u003e\u003cp\u003eIn contrast, the genetic variant rs1441754 near the lipoprotein lipase (\u003cem\u003eLPL\u003c/em\u003e) gene, previously identified as a cholesterol-associated variant, exhibited significant allele-specific activity in LX-2 but showed far less differential activity in HepG2 \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eM-\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eN\u003cb\u003e)\u003c/b\u003e\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e. Luciferase reporter assays confirmed this cell type-specific enhancer activity, with significant allele-specific effects observed exclusively in LX-2 cells, but not in HepG2 or Huh7 hepatocyte cell lines \u003cb\u003e(Extended Data Fig.\u0026nbsp;14J)\u003c/b\u003e. Interestingly, the predicted target gene of rs1441754, \u003cem\u003eLPL\u003c/em\u003e, is upregulated in hepatic stellate cells and exacerbates liver fibrosis in MASLD in mouse models\u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e. Mechanistically, we found that rs1441754 resides in stellate-specific chromatin peaks and may regulate \u003cem\u003eLPL\u003c/em\u003e expression by affecting the binding motif of RUNX \u003cb\u003e(Extended Data Fig.\u0026nbsp;14K)\u003c/b\u003e. Strikingly, deletion of the rs1441754-containing region in LX-2 resulted in a significant increase in \u003cem\u003eLPL\u003c/em\u003e expression, suggesting its role as a negative regulatory element in stellate cells \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eO\u003cb\u003e)\u003c/b\u003e. Functionally, rs1441754 deletion also resulted in the upregulation of multiple fibrosis-associated genes in LX-2 cells \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eO-\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eP\u003cb\u003e)\u003c/b\u003e. Together, these findings highlight the integrative roles of context-specific DAVs in modulating key regulators of triglyceride metabolism including APOA5, ANGPTL3, and LPL, and their downstream contributions to altered lipid homeostasis and hepatic stellate cell activation in MASLD \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eQ\u003cb\u003e)\u003c/b\u003e.\u003c/p\u003e\u003cp\u003eIn addition to the role of rs1441754 in stellate cell activation, we identified another regulatory variant, rs66477683, which modulates \u003cem\u003eMERTK\u003c/em\u003e expression \u003cb\u003e(Extended Data Fig.\u0026nbsp;15A-15B)\u003c/b\u003e. Given prior reports linking MERTK to liver fibrosis\u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e, our findings suggest that rs66477683 may represent a fibrosis-driving variant acting through transcriptional control of \u003cem\u003eMERTK\u003c/em\u003e \u003cb\u003e(Extended Data Fig.\u0026nbsp;15C-15D)\u003c/b\u003e.\u003c/p\u003e\u003cp\u003e\u003cb\u003eFunctional linking of rs474513 to\u003c/b\u003e \u003cb\u003eSLC22A3\u003c/b\u003e \u003cb\u003eand hepatic lipid homeostasis\u003c/b\u003e\u003c/p\u003e\u003cp\u003eNotably, DAV rs474513, previously linked to obesity and type 2 diabetes\u003csup\u003e\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e, is located in hepatocyte-specific open chromatin regions near the promoter of the \u003cem\u003eSLC22A3\u003c/em\u003e gene. Chromatin interaction data also revealed a strong enhancer-promoter interaction between the genomic region containing rs474513 and the \u003cem\u003eLPA\u003c/em\u003e gene, suggesting the regulatory potential of rs474513 on \u003cem\u003eSLC22A3\u003c/em\u003e and \u003cem\u003eLPA\u003c/em\u003e\u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eA\u003cb\u003e)\u003c/b\u003e. Interestingly, public liver eQTL data showed that the genotype of rs474513 is closely associated with the expression levels of both \u003cem\u003eSLC22A3\u003c/em\u003e and \u003cem\u003eLPA\u003c/em\u003e, but in opposite directions \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eB\u003cb\u003e)\u003c/b\u003e. Livers carrying the alternative allele G exhibited higher expression of \u003cem\u003eSLC22A3\u003c/em\u003e and lower expression of \u003cem\u003eLPA\u003c/em\u003e compared to those carrying the reference allele A. Mechanistically, rs474513 modulates the binding motif for RUNX, a transcription factor known to be involved in MASLD progression\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e. The agreement score with the RUNX consensus motif was 997-fold greater for the G allele than for the A allele \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eC\u003cb\u003e)\u003c/b\u003e. Consistently, overexpression of RUNX1/2 increased the activity of reporters carrying the G allele compared to those with the A allele \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eD\u003cb\u003e)\u003c/b\u003e. In contrast, inhibition of RUNX2 activity using small molecule CADD522 reduced \u003cem\u003eSLC22A3\u003c/em\u003e expression, but had no effect in HepG2 cell where a small genomic region containing rs474513 was deleted \u003cb\u003e(Extended Data Fig.\u0026nbsp;16A-16B)\u003c/b\u003e. Moreover, deletion of the genomic region containing rs474513 led to decreased \u003cem\u003eSLC22A3\u003c/em\u003e expression and increased \u003cem\u003eLPA\u003c/em\u003e expression, in line with the eQTL data \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eE\u003cb\u003e)\u003c/b\u003e. Similar gene expression patterns were observed when the genomic region was inactivated using CRISPRi technology \u003cb\u003e(Extended Data Fig.\u0026nbsp;16C)\u003c/b\u003e. Importantly, the G\u0026thinsp;\u0026gt;\u0026thinsp;A-edited heterozygous HepG2 cells also exhibited decreased \u003cem\u003eSLC22A3\u003c/em\u003e expression \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eF \u003cb\u003eand Extended Data Fig.\u0026nbsp;16D)\u003c/b\u003e.\u003c/p\u003e\u003cp\u003eTo examine the role of rs474513 in lipid metabolism associated with MASLD, we treated the rs474513-deleted or gene-edited HepG2 with PAOA to model steatosis. Interestingly, we found that lipid accumulation was increased in rs474513-deleted and G\u0026thinsp;\u0026gt;\u0026thinsp;A-edited HepG2 cells treated with PAOA \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eG-\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eJ\u003cb\u003e)\u003c/b\u003e. Given the established role of RUNX in mediating the transcriptional activity of rs474513, we next treated HepG2 with the RUNX1/2 inhibitors Ro5-3335 and CADD522. Pharmacologic inhibition of RUNX signaling led to reduced \u003cem\u003eSLC22A3\u003c/em\u003e expression and further exacerbated lipid accumulation, mirroring the effects of rs474513 deletion \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eK \u003cb\u003eand Extended Data Fig.\u0026nbsp;16E-16F)\u003c/b\u003e. Importantly, ectopic overexpression of SLC22A3 rescued lipid accumulation phenotype \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eK \u003cb\u003eand Extended Data Fig.\u0026nbsp;16E-16F)\u003c/b\u003e, confirming that SLC22A3 acts downstream of the rs474513-RUNX regulatory axis to maintain hepatic lipid homeostasis.\u003c/p\u003e\u003cp\u003eTo further investigate the physiological function of rs474513 in MASLD-relevant organoid model\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e, we deleted a small genomic region containing rs474513 in human induced pluripotent stem cell (hiPSC) line and observed a reduction in \u003cem\u003eSLC22A3\u003c/em\u003e expression \u003cb\u003e(Extended Data Fig.\u0026nbsp;16G-16H)\u003c/b\u003e, consistent with the above finding. Importantly, the deletion of rs474513 did not impair the efficiency of hepatic differentiation in the hiPSC line \u003cb\u003e(Extended Data Fig.\u0026nbsp;16I)\u003c/b\u003e. We then differentiated both the wild-type and rs474513-depleted hiPSC lines into human liver organoids (HLOs) and exposed them to free fatty acids to model the steatosis process\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e. Interestingly, the rs474513-depleted HLOs exhibited higher lipid accumulation compared to wild-type HLOs, along with increased expression of \u003cem\u003ePLIN2\u003c/em\u003e, a gene involved in lipid storage and metabolism \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eL-\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eM\u003cb\u003e)\u003c/b\u003e. We also observed elevated levels of \u003cem\u003eCD68\u003c/em\u003e and \u003cem\u003eCK7\u003c/em\u003e in rs474513-depleted HLOs \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eL-\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eM\u003cb\u003e)\u003c/b\u003e, indicating an enhanced inflammatory response and ductal reaction. Furthermore, the higher ACTA2 and lower ALBUMIN expression implied the potential risk of fibrosis.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eFunctional DAVs improve MASLD risk prediction\u003c/h3\u003e\n\u003cp\u003ePolygenic risk scores (PRS) aggregate the effects of multiple genetic variants to estimate an individual's likelihood of developing a disease\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e. To assess whether hepatic DAVs can effectively predict MASLD risk, we compared the performance of the PRS constructed from 114 LD-independent MPRA-prioritized variants (DAVs) \u003cb\u003e(Table \u003cspan refid=\"MOESM7\" class=\"InternalRef\"\u003eS7\u003c/span\u003e)\u003c/b\u003e to the control PRS (MPRA-Ctrl) constructed from randomly selected, matched set of MPRA total variants using data from the Million Veteran Program (MVP)\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e. Our analysis revealed that hepatic DAVs PRS outperformed the control PRS in predicting various MASLD-related traits, including MASLD defined by chronic elevation of liver enzyme alanine aminotransferase (MASLD-ALT), MASLD defined by international classification of disease (ICD) codes (MASLD.ICD), ALT levels, and fibrosis index FIB4 progression \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eN\u003cb\u003e)\u003c/b\u003e. Notably, the individuals with DAV PRS scores in the top 10% exhibited a 1.6-fold higher risk of developing MASLD compared to the ones in the lowest 10%, which is better than the performance of MPRA-Ctrl PRS, and comparable to the PRS built from nine well-established missense variants contributing to MASLD (PRS9), including the rs738409 (I148M) variant of the gene \u003cem\u003ePNPLA3\u003c/em\u003e and rs58542926 (E167K) of the gene \u003cem\u003eTM6SF2\u003c/em\u003e \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eO \u003cb\u003eand Table \u003cspan refid=\"MOESM8\" class=\"InternalRef\"\u003eS8\u003c/span\u003e)\u003c/b\u003e\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. We further benchmarked the DAV-PRS in the \u003cem\u003eAll of Us Research Program\u003c/em\u003e against previously published MASLD PRS models and found that, while our MPRA-prioritized PRS performed similarly to PRS9, its predictive power was modestly lower than PGS002732, the best-performing published model \u003cb\u003e(Extended Data Fig.\u0026nbsp;16J)\u003c/b\u003e\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. These data underscore the power of our approach in identifying functional variants linked to MASLD, which can also be extended to other complex diseases.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eFine-mapping causal variants for diseases or complex traits continues to be a major hurdle in the field of genetics\u003csup\u003e\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u003c/sup\u003e. Numerous computational algorithms have been developed, employing various statistical and computational strategies to fine-map genetic variants. For example, FINEMAP and DAP utilize Bayesian frameworks to estimate posterior probabilities of causality; PAINTOR and fGWAS incorporate functional annotations to prioritize variants; Coloc and eCAVIAR identify shared causal variants between complex traits\u003csup\u003e\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e. Additionally, deltaSVM enhances fine-mapping by providing functional predictions for non-coding variants that disrupt transcription factor binding using machine learning techniques\u003csup\u003e\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u003c/sup\u003e. However, computational prediction algorithms for prioritizing functional variants are often not concordant among themselves and are frequently inconsistent with experimental validation\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. Here, by integration of snATAC, high-throughput MPRA, and CROP-seq data, we identified functional causal variants and revealed key disease-associated cell types or cell states and elucidated the transcriptional programs influenced by these functional variants. Remarkably, functional variants identified in this way regulate hepatic lipid metabolism and stellate cells activation and effectively predict MASLD risk.\u003c/p\u003e\u003cp\u003eTo our knowledge, this is the first study to comprehensively profile the heterogeneity of chromatin landscapes in individuals at different stages of MASLD and thus provides a high-quality human liver regulome resource. Integrating GWAS variants with single-cell chromatin landscape data revealed that MASLD risk variants are enriched in disease-specific hepatocyte and stellate cell states. Moreover, by combining motif activity and risk variant enrichment analyses at the single-cell level, we demonstrated how these MASLD risk variants regulate gene expression. Furthermore, the single-cell chromatin accessibility data enabled us to nominate target genes of risk variants using co-accessibility analysis. Notably, snATAC-seq studies using large cohorts of MASLD liver samples will enhance the statistical power to pinpoint the functional variants by performing chromatin accessibility QTL (caQTL) or allele-imbalance analyses\u003csup\u003e\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eImportantly, our MPRAs were conducted using two disease-relevant cell lines under MASLD-associated environmental conditions, demonstrating that these DAVs exhibit cell type-specific and stimulus-dependent activity. This highlights the importance of selecting appropriate cell lines and cellular contexts for identifying functional SNPs. Notably, only 64% of DAVs exhibited concordant effect directions between reporter activity and eQTL effect size, while the remaining 36% showed discordance, likely reflecting the cellular heterogeneity of liver tissue and the context-dependent nature of chromatin environments. Future implementation of this approach in human liver organoids or \u003cem\u003ein vivo\u003c/em\u003e models may provide more physiologically relevant insights\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. Moreover, using various CRISPR approaches, we identified MASLD-associated transcriptional programs. We observed that rs474513 affects the expression of \u003cem\u003eSLC22A3\u003c/em\u003e and \u003cem\u003eLPA\u003c/em\u003e in opposite directions, demonstrating how a single variant can disparately regulate multiple genes\u003csup\u003e\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e. Conversely, both rs1748197 and rs10889356 regulate the expression of \u003cem\u003eANGPTL3\u003c/em\u003e, suggesting synergistic effects of these genetic variants. Remarkably, we identified several MASLD-associated DAVs that regulate the core regulatory network of LPL activity, including APOA5 (an LPL activator), ANGPTL3 (an LPL inhibitor), and LPL itself. These findings suggest that MASLD development is possibly mediated by local paracrine effects on LPL activity in the liver. Importantly, using physiologically relevant human liver organoids and gene-editing tools, we demonstrated the causal roles of these genetic variants in lipid metabolism associated with MASLD.\u003c/p\u003e\u003cp\u003eIn sum, our study combined high-throughput MPRA, snATAC-seq, and CROP-seq to identify functional causal variants and their target genes in MASLD-related cell types. This broadens our understanding of genetic mechanisms underlying MASLD risk, and will be a valuable guide to prediction and treatment of MASLD and related disorders. Moreover, this approach may be employed to better understand other complex human diseases.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank all other members of the Hu and Lazar labs for their invaluable help and discussion. We would also like to thank the Liver Tissue Cell Distribution System and SEKISUI XenoTech for providing the MASLD and healthy liver samples. We thank Shixiang Wang from Dr. Yun Shen Chan lab for the help with the hepatic organoid differentiation. We thank Dr. Yajie Zhao for the discussion on PRS analysis. This PRS analysis is based on data from the Million Veteran Program (MVP), Office of Research and Development, Veterans Health Administration with grateful support by award MVP000 and Veterans who participated in MVP. This publication does not represent the views of the Department of Veterans Affairs, the US Food and Drug Administration or the US Government. This work was supported by the National Key R\u0026amp;D Program of China (2024YFF0509000 to W.H.), the National Natural Science Foundation of China (82270868 to W.H.), the R\u0026amp;D Program of Guangzhou National Laboratory (GZNL2023A02004, GZNL2023A03004, and GZNL2023A02002 to W.H.), the Pearl River Talent Recruitment Program (2021QN02Y963 to W.H.), National Institutes of Health grant (R01 DK125573, R01DK49780, and P30 DK019525 to M.A.L., R01 DK134575 to V.M.), the JPB Foundation (to M.A.L.), the Cox Medical Institute (to M.A.L.), VA Merit Funding BX003362 from VA Office of R\u0026amp;D (K.M.C.), and VA Informatics and Computing Infrastructure (VINCI) VA HSR RES 130457 (J.A.L.). Y.X. was supported by National Institutes of Health grant (K01 DK138281) and American Heart Association Training grant 827529. B.Z. was supported by Guangzhou Municipal Science and Technology Bureau SL2022A04J00542.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eW.H., Y.X. and M.A.L. conceptualized the study, interpreted data and wrote the manuscript, which was revised and approved by all authors. B.Z., and B.C. performed MPRA and functional validation experiments. Y.X., Y.L., and J.Z. performed snATAC-seq.\u0026nbsp;C.L. and N.H. performed bioinformatics analysis. \u0026nbsp; X.C and F.Y. help with GWAS risks enrichment analysis. M.V., J.A.L., K.C., R.M., B.P. and D.J.R. performed PRS analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe accession number for the sequencing data reported in this paper are GSE281367, GSE281364, and GSE281160.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare the following competing interests: MAL is on a scientific advisory board of Pfizer and is co-founder and serves on the scientific advisory board of Flare Therapeutics. DJR is on the scientific advisory board for Alnylam, Novartis, and Verve. \u0026nbsp;All other authors have no competing interests to declare.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eCell lines and culture\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHuman hepatoma cell line HepG2 and Huh-7, human hepatic stellate cell line LX-2, and human embryonic kidney (HEK) 293T cell line were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum (FBS). Cells were maintained in a humidified incubator at 37\u0026nbsp;°C with 5% CO₂and supplemented with 100 U/mL penicillin and 100 µg/mL streptomycin. To establish an \u003cem\u003ein vitro\u003c/em\u003e lipotoxic model, HepG2 or Huh-7 cells were treated with 0.15 mM palmitic acid (PA) and 0.3 mM oleic acid (OA). Both fatty acids were dissolved in 0.5% fatty acid-free bovine serum albumin (BSA). The treatment was administered for 24 h. To establish an \u003cem\u003ein vitro\u003c/em\u003e fibrosis model, LX-2 cells were treated with 20 ng/mL transforming growth factor-beta 1 (TGFβ\u003csub\u003e1\u003c/sub\u003e) for 48 h.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHuman iPSCs culture\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHuman induced pluripotent stem cell (iPSC) lines were generated following the study protocol approved by The University of Pennsylvania Human Subjects Research Institutional Review Board, as previously reported\u003csup\u003e66\u003c/sup\u003e. iPSCs were cultured in mTeSR™ medium at 37 °C with 5% CO₂. Gene-edited stable monoclonal iPSC lines were produced as previously described\u003csup\u003e67\u003c/sup\u003e. In brief, human iPSCs were infected with Exci-sgRNA lentivirus co-expressing Cas9 and selected for 3 days using puromycin (0.5 µg/ml) and blasticidin (5 µg/ml). The surviving iPSCs were then subcloned at low seeding density and cultured for one week. Following colony isolation and expansion, Sanger sequencing was performed to identify stable DAV-deleted monoclonal human iPSC lines.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHuman studies\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNormal and MASLD human livers for snATAC-seq were provided by the Liver Tissue Cell Distribution System, Minneapolis, Minnesota, which was funded by National Institutes of Health contract no. HSN276201200017C. Additional human livers were obtained from SEKISUI XenoTech. All studies using human samples were exempted by the University of Pennsylvania Institutional Review Board.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHepG2 CRISPRi/a cell lines\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;construction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe CRISPRi_HepG2 cell line was generated by targeted integration of a UCOE-CAG promoter-driven dCas9-KRAB/MeCP2-T2A-BSD expression cassette into the AAVS1 \"safe harbor\" site using CRISPR/Cas9. Donor plasmid (pUC19-AAVS1-dCas9-KRAB/MeCP2) and AAVS1 targeting plasmid (PX330-sgAAVS1_human) were co-transfected into HepG2 cells using Lipofectamine 3000 reagent. Blasticidin selection (30 µg/mL) commenced 72 h post-transfection and was maintained for at least 7 days. Expression of dCas9-KRAB/MeCP2 was evaluated by immunofluorescence.\u003c/p\u003e\n\u003cp\u003eThe CRISPRa_HepG2 cell line was established by infecting HepG2 cells with lentiviruses containing a reverse tetracycline-controlled trans-activator (rtTA) expression cassette co-expressing blasticidin S deaminase (BSD) and a 3G-Tet-regulated dCas9-VPR expression cassette. Blasticidin selection (30 µg/mL) was initiated 72 h post-infection and maintained for at least 7 days. Expression of dCas9-VPR was induced by treating cells with doxycycline (5 µg/mL) for 3 days and subsequently evaluated by immunofluorescence.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePlasmids construction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe donor plasmid (pUC19-AAVS1-dCas9-KRAB/MeCP2) was generated by cloning a UCOE-CAG-dCas9-KRAB/MeCP2-T2A-BSD expression cassette alongside with AAVS1 homology arm (HA) into pUC19 backbone. The HA-L, HA-R and UCOE element were amplified from human genome. The CAG promoter was amplified from SMARTvector. The dCas9-KRAB/MeCP2, T2A-BSD-WPRE and SV40-polyA fragments were amplified from the pLX-TRE-dCas9-KRAB/MeCP2-BSD vector. For PX330-sgAAVS1 plasmid, annealed oligo duplex (sgAAVS1_human) was prepared and cloned into BbsI cut pX330-U6-hSpCas9 backbone.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe TetON-dCas9-VPR-BSD plasmid was generated by replacing the KRAB/MeCP2 expression sequence in the pLX-TRE-dCas9-KRAB/MeCP2-BSD vector with VPR expression sequence amplified from SP-dCas9-VPR plasmid. All the primers are shown in Supplemental Table S9. To produce lentiviruses, the lentivirus expression vector and the packaging plasmid mix (psPAX2 and pMD2.G) were co-transfected into HEK-293T cells. The lentivirus supernatant was harvested and stored in aliquots at -80°C until use.\u003c/p\u003e\n\u003cp\u003eFor pcDNA3.1-\u003cem\u003eSLC22A3\u003c/em\u003e overexpression (OE) plasmid construction. The human \u003cem\u003eSLC22A3\u003c/em\u003e CDS gene was obtained using HepG2 cDNA as a template by PCR amplification. The expression plasmid vector (pcDNA3.1+) was linearized by NheI and KpnI (NEB, Beijing, China) restriction enzyme digestion. The pcDNA3.1-\u003cem\u003eSLC22A3\u003c/em\u003e vector was constructed according to ClonExpress Ultra One Step Cloning Kit.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLiver processing and nuclei isolation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIsolation of nuclei for snATAC-seq was performed as previously described \u003csup\u003e17\u003c/sup\u003e. Briefly, frozen human liver tissue was homogenized using a Dounce homogenizer in lysis dilution buffer containing 10 mM Tris-HCl (pH 7.5), 10 mM NaCl, 3 mM MgCl₂, 1% BSA, and cOmplete™ Protease Inhibitor Cocktail. The homogenate was centrifuged, and the pellet was resuspended in lysis dilution buffer. To the nuclei suspension, lysis buffer (10 mM Tris-HCl, pH 7.5; 10 mM NaCl; 3 mM MgCl\u003csub\u003e2\u003c/sub\u003e; 1% BSA; 0.1% Tween-20; 0.1% NP-40; 0.01% digitonin; cOmplete™ Protease Inhibitor Cocktail) was added. The nuclei were washed and resuspended in lysis dilution buffer. Nuclei were further purified using the Nuclei Isolation Kit with a 1.8 M sucrose cushion, following the manufacturer's instructions. After centrifugation at 30,000 g for 45 minutes at 4 °C, the nuclei pellet was collected and resuspended in PBS containing 1% BSA. The nuclei were filtered through a 40 μm cell strainer before loading onto the 10X\u0026nbsp;Genomics Chromium Controller.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003esnATAC-seq libraries construction and sequencing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003esnATAC-seq libraries were prepared according to the manufacturer’s instructions of 10XGenomics (PN-1000176). Libraries were sequenced with Novogene (https://www.novogene.com/us-en/). 400 million paired-end reads were obtained for each sample using the sequencing parameters recommended in 10X Genomics guide.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCluster analysis of snATAC\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003esnATAC-seq data were preprocessed using the 10X Genomics pipeline (Cell Ranger ATAC v2.1.0) aligned to the hg38 genome assembly with default parameters. Samples from different patients were processed separately. High-quality individual cells were retained for downstream analysis (1,000 \u0026lt; unique barcode number \u0026lt; 6,000; reads in peaks percent \u0026gt; 15%; blacklist ratio \u0026lt; 0.05; TSS enrichment \u0026gt; 3; nucleosome signal \u0026lt; 4). Quality control measurements and filtering were conducted using Signac (v. 1.12.9003)\u003csup\u003e68\u003c/sup\u003e. After quality control, all samples were merged into a single Seurat object. The integration function in Signac was then employed to remove batch effects. Unsupervised clustering was performed at a resolution of 0.4. Cell types were identified by integrating liver snRNA data with groups determined through unsupervised clustering. Motif activity was assessed using chromVAR (v. 1.24.0)\u003csup\u003e69\u003c/sup\u003e, and pseudotime scores were calculated with Monocle3 (v. 1.3.4). The pseudotime analysis was further confirmed using StaVia (v. 2.0). Gene Ontology (GO) analysis was conducted using GREAT (v. 4.0.4)\u003csup\u003e70\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIntegrative analysis of snATAC-seq and snRNA-seq\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe liver snRNA data used for the integrative analysis was downloaded from GSE189600 \u003csup\u003e17\u003c/sup\u003e. Cells were labeled according to the cell type information provided in the original publication. Subsequently, cells in the liver snATAC dataset were annotated by transferring cell labels from the liver snRNA dataset using tools from the Seurat (v. 5.0.2) package\u003csup\u003e71\u003c/sup\u003e. Utilizing the transferred labels and cell types identified through cluster-specific gene scores, correlation analyses were performed to assess the concordance between chromatin accessibility and gene expression profiles.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCell type-specific peaks and gene scores\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCell type-specific peaks were identified by selecting significant peaks (adjusted p-value \u0026lt; 0.05; log₂ fold change \u0026gt; 0.58) that were not significant in more than three other cell types. Gene scores were calculated using the GeneActivity function in Signac (v1.12.9003)\u003csup\u003e68\u003c/sup\u003e, which quantifies the number of fragments per cell mapping to both the gene body and the promoter region (2 kb upstream).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGWAS enrichment analysis by LDSC and SCAVENGE\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe LDSC package (\u003ca href=\"https://github.com/bulik/ldsc\" target=\"_new\"\u003ehttps://github.com/bulik/ldsc\u003c/a\u003e) was used to perform LDSC analysis. GWAS summary statistics were downloaded from the GWAS Catalog: Triglycerides (GCST90092955), BMI (GCST90179150), and MASLD (GCST90054782). These GWAS summary statistics were converted to a format compatible with LDSC \u003csup\u003e72\u003c/sup\u003e using the provided script (munge_sumstats.py). For each cell type-specific peak, bed files were lifted over to hg19 from hg38 to create annotation files for each cell type. Coefficient and enrichment calculations were performed by LDSC following the cell-type-specific analysis tutorial (\u003ca href=\"https://github.com/bulik/ldsc/wiki/Cell-type-specific-analyses\" target=\"_new\"\u003ehttps://github.com/bulik/ldsc/wiki/Cell-type-specific-analyses\u003c/a\u003e).\u003c/p\u003e\n\u003cp\u003eSCAVENGE analysis were perform as described\u003csup\u003e20\u003c/sup\u003e, GWAS summary statistics were downloaded from the GWAS Catalog for MASLD (GCST90091033), liver fat (GCST90016673) and cirrhosis (GCST90319878). Sentinel SNPs were identified using GCTA (v. 1.94.1)\u003csup\u003e73\u003c/sup\u003e with a significance threshold of p \u0026lt; 10\u003csup\u003e−8\u003c/sup\u003e. Fine-mapping analysis was conducted using coloc (v. 5.2.3)\u003csup\u003e74\u003c/sup\u003e with adj.P-value cutoff of 0.0001. The output from coloc was lifted over to hg19 from hg38 for downstream analysis. Information on counts, UMAP embeddings, and LSI embeddings was extracted from the results of the Signac analysis. Subsequently, SCAVENGE (v. 1.0.2)\u003csup\u003e20\u003c/sup\u003e utilized these data to compute the trait-relevant score (TRS).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eJoint analysis of single cell motif enrichment and GWAS enrichment\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJoint analysis of single-cell motif and GWAS enrichments was performed as previously described\u003csup\u003e10\u003c/sup\u003e, with minor modifications. Briefly, TF motif activity for each cell was quantified using chromVAR\u003csup\u003e69\u003c/sup\u003e, which infers TF activity by quantifying deviations in chromatin accessibility across sets of peaks sharing the same TF-binding motif. ChromVAR calculates a bias-corrected deviation score and corresponding Z-score by comparing the observed accessibility of motif-containing peaks to an expected accessibility derived from matched background peaks. \u0026nbsp;Trait relevance scores (TRS) for individual cells were computed using SCAVENGE\u003csup\u003e20\u003c/sup\u003e. Spearman’s rank correlation coefficients were then calculated between motif activity and TRS across all cells to assess the association between transcription factor activity and GWAS trait enrichment.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMPRA library design and construction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMASLD-associated variants were obtained from the GWAS Catalog and relevant publications (Table S1). SNPs in Linkage disequilibrium (LD) with all collected significant SNPs were identified using the LDProxy function from LDlinkR, applying thresholds of minor allele frequency (MAF) \u0026gt; 5% and r² \u0026gt; 0.8 in European populations. Variants located within coding regions, as well as indels and multiallelic variants, were excluded from the analysis. For each remaining variant, a 126-base pair (bp) genomic sequence centered on the reference or alternate allele was extracted for oligonucleotide synthesis. In total, 5,369 variants were included for the MPRA library construction.\u003c/p\u003e\n\u003cp\u003eThe oligonucleotide library was designed and synthetized as 170 bp oligo pool, comprising a 126 bp genomic sequence (candidate regulatory sequence, CRS) flanked by 22 bp adapter sequences at both ends (5’-GAGTACTGTATGGGCGGGTACC [126 bp oligo] GCTAGCGACAAAAGTGTCAACT-3’). The MPRA plasmid library was constructed using the pGL4.24-miniP-GFP vector, where the luciferase expression sequence in the original pGL4.24 vector was replaced with a green fluorescent protein (GFP) expression sequence. A two-step PCR was employed to incorporate a minimal TATA promoter (miniP), unique 20 bp barcodes, and the necessary vector overhang sequences into the oligonucleotide library. The oligo pool PCR products were purified and subsequently cloned into the KpnI and NcoI digested pGL4.24-miniP-GFP vector using the ClonExpress Ultra One Step Cloning Kit as described above. The ligation reactions were purified and transformed into Lucigen Endura Electrocompetent Cells (Biosearch Technologies, 60242-2) via electroporation. The resulting plasmid library was purified using the ZymoPURE II™\u0026nbsp;Plasmid Maxiprep Kit. Finally, CRS-barcode fragments were amplified by PCR using the primers CRS/BC_F and CRS/BC_R for subsequent sequencing to establish CRS/barcode pairings.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMassively parallel reporter assay\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe MPRA libraries were transfected into cells using Lipofectamine 3000 reagent. For HepG2 MPRA experiments, 2 × 10⁷ cells were used for one replicate. 16 h post-transfection, the medium was refreshed with complete culture medium with or without PAOA treatment and incubated for another 24 h. For LX-2 MPRA experiments, 1 × 10⁷ cells were used for one replicate. 6 h post-transfection, the medium was refreshed with complete culture medium with or without TGFβ treatment and incubated for an additional 36 h. A total of four biological replicates were performed for each experiment.\u003c/p\u003e\n\u003cp\u003eCells were harvested and performed DNA and RNA extraction using the Quick DNA/RNA Miniprep Kit following the manufacturer’s protocol. RNA samples are rigorous DNase treated using the RNase-free Recombinant DNase I Set. For each biological replicate, 240 µg of RNA was used to synthesize the first-strand cDNA library using SuperScript II in combination with a GFP mRNA-specific primer (MPRA_BCshort_R). To amplify random barcode from plasmid DNA or cDNA, and incorporate required P5/P7 flowcell sequence and sample index, the first-round PCR was performed using the primers MPRA_BClong_P5i5_F and MPRA_BClong_P7i7_F, followed by the second-round PCR using the primers P5 and P7. Each PCR reaction utilized 48 µg of DNA or total cDNA per biological replicate as the template. All PCR reactions were carried out using the NEBNext® High-Fidelity 2X PCR Master Mix. DNA and RNA barcode libraries were pooled in a 1:3 mole ratio. Library representation was assessed by next-generation sequencing on Illumina Novaseq platform.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMPRA data analysis\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMPRA plasmid library Read 1 sequences were processed by trimming the fixed sequence preceding the inserted oligo sequences using cutadapt (https://doi.org/10.14806/ej.17.1.200) with the parameters --discard-untrimmed -m 80 -g GGCCTAACTGGCCGGTACCCTGAGTACTGTATGGGCGGGTACC, retaining only the subsequent sequences corresponding to the synthesized oligos and discarding reads shorter than 80 bp. Similarly, Read 2 sequences were trimmed to remove the fixed sequence preceding the inserted barcode using cutadapt with parameters --discard-untrimmed -l 20 -g TCACCATGGTGGCTTTACCAACAG, extracting the following 20 bp as barcode sequences. The first 107 bp of each synthesized oligo were extracted to create a reference genome, and Read 1 sequences were mapped to this reference using bwa mem\u003csup\u003e75\u003c/sup\u003e with parameters -L 100 -k 8 -O 5. Reads with a mapping score below 0.95 (mapping score = matched bases / total bases) were removed. Paired Read 2 sequences from the plasmid library corresponding to the retained Read 1 were used to construct oligo-barcode pairs, with duplicate pairs and barcodes mapping to multiple oligos further filtered to ensure unique pairings. For both cDNA and DNA libraries, fixed sequences were trimmed from the beginning of Read1 using cutadapt with parameters --discard-untrimmed -l 20 -g CTTGGCAATCCGGTACTTAGCGCATCTAAC. The reverse complement of the subsequent 20 bp sequences was treated as barcode sequences, and only barcodes present in the previously identified unique oligo-barcode pairs were retained for downstream analysis.\u003c/p\u003e\n\u003cp\u003eBoth DNA and cDNA barcodes in each oligo-barcode pair were quantified, and barcodes with a DNA barcode count of zero were excluded to ensure data quality. To assess oligo activity, we modified the MPRAprofiler protocol as described \u003csup\u003e76\u003c/sup\u003e, summarizing the total barcodes associated with each oligo for subsequent analyses. Potential enhancers were identified using DESeq2\u003csup\u003e77\u003c/sup\u003e to compare cDNA and DNA barcode counts for each oligo, with oligos exhibiting an adjusted p-value \u0026lt; 0.05 and a log₂ fold change \u0026gt; 0 (cDNA/DNA) classified as enhancers. Variants with either allele identified as active enhancers were further analyzed to determine differences in regulatory activity between the reference and alternative alleles. To detect differentially active variants, we first calculated the count ratio between the reference and alternative alleles for all variants. Subsequently, using the median ratio of four DNA replicates as a control, we normalized all RNA ratios and performed a Student’s \u003cem\u003et\u003c/em\u003e-test to calculate p-values. The Benjamini-Hochberg method was applied to adjust the p-values for multiple testing, and variants with an adjusted p-value \u0026lt; 0.01 were designated as significant differentially active variants.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDAVs functional annotations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDAVs were annotated to different genomic regions using ChIPseeker\u003csup\u003e78\u003c/sup\u003e. Additionally, published liver chromatin interaction data, eQTLs, and ChromHMM data were downloaded to elucidate the functional significance of DAVs. Overlaps between DAVs and chromatin loops were identified using the following datasets: (1) right lobe liver intact Hi-C data from ENCODE\u003csup\u003e79\u003c/sup\u003e (ID ENCFF961GKA); (2) HepG2 intact Hi-C data from ENCODE (ID ENCFF264RQT); (3) HepG2 POLR2A ChIA-PET data from ENCODE (ID ENCSR857MYZ); (4) capture-C data of promoters and H3K4me3/H3K27ac peaks in liver HepG2 cells\u003csup\u003e33\u003c/sup\u003e; and (5) liver promoter capture Hi-C data from GSE86189\u003csup\u003e34\u003c/sup\u003e, including promoter-other interactions. Each loop anchor was extended by a fixed 2.5 kb on either side. DAVs with significant variant-gene pairs were annotated using liver eQTL data from GTEx v8\u003csup\u003e80\u003c/sup\u003e. Chromatin states of DAVs were predicted using the core 15-state model for human liver-related tissues and cells, as determined by ChromHMM\u003csup\u003e81\u003c/sup\u003e and downloaded from the ROADMAP Epigenetics Project\u003csup\u003e82\u003c/sup\u003e (IDs E066, E118). Eleven of these states were considered active and used for DAV annotation: 1_TssA, 2_TssAFlank, 3_TxFlnk, 4_Tx, 5_TxWk, 6_EnhG, 7_Enh, 8_ZNF/Rpts, 10_TssBiv, 11_BivFlnk, and 12_EnhBiv. The Bedtools\u003csup\u003e83\u003c/sup\u003e intersect command was utilized to calculate the overlap between the centers of DAVs and loop anchors, eQTLs, and ChromHMM-defined chromatin states.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEnrichment of DAVs in cell type peaks\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo determine whether DAVs are preferentially associated with specific cell types, we assessed the overlap between the DAVs and cell type-specific peaks derived from scATAC-seq. This analysis involved comparing DAV overlaps with cell type-specific peaks against controls generated from 1,000 iterations of randomly selected genomic regions. Fisher’s exact test was employed to calculate p-values, determining the statistical significance of the observed overlaps. Additionally, we compared the number of overlaps between DAVs and all tested variants within specific cell type peaks to evaluate the enrichment tendency of DAVs relative to non-functional variants.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe also analyzed the enrichment of all SNPs assayed by MPRA across promoter and enhancer elements in different cell types. Peaks located within promoter-proximal regions (-2 kb to +0.5 kb relative to transcription start sites) were classified as promoters, while all other peaks were designated as enhancers.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTranscriptional factor binding analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTranscription factor binding site (TFBS) disruptions by DAVs were assessed using SNP2TFBS and motifBreakR\u003csup\u003e84,85\u003c/sup\u003e. In the SNP2TFBS analysis, statistical significance was determined by comparing the proportion of DAVs predicted to disrupt a given TF motif to the proportion of background SNPs affecting the same motif. In contrast, the motifbreakR approach assessed significance by comparing the proportion of DAVs disrupting a TF motif to the proportion of all MPRA-tested SNPs with predicted effects on that TF. TFBS with an adjusted p-value \u0026lt; 0.01 were considered significantly affected by DAVs. Additionally, a more permissive approach was implemented to identify potentially disrupted TFBS by individual DAVs. This involved extracting the alternative and reference sequences centered on each variant, spanning ± 20 bp (totaling 41 bp). Motifs within these variant sequences were identified using MEME’s FIMO\u003csup\u003e86\u003c/sup\u003e against the HOCOMOCOv11_core_HUMAN database and HOMER TF database\u003csup\u003e87\u003c/sup\u003e, retaining only those motifs that encompassed the SNP location regardless of strand orientation. Subsequently, the position frequency matrix (PFM) values at the SNP position were compared between the reference and alternative alleles. Motifs with either the reference or alternative PFM value exceeding 0.3 and a ratio greater than 2 were designated as potentially interrupted motifs. For the exploration of differentially enriched TF motifs in DAVs and enhancers across various cell types and states, HOMER was utilized to identify motifs within condition-specific enhancers and DAVs, using other condition-specific enhancers or DAVs as background controls. Only motifs with a q-value \u0026lt; 0.1 were deemed enriched in DAVs and enhancers.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFor the significantly disrupted TFs by DAVs either in HepG2 or LX-2 cells, under both control and PAOA or TGFβ\u0026nbsp;treatment, we constructed protein-protein interaction network by STRING\u003csup\u003e88\u003c/sup\u003e (https://cn.string-db.org/) with default parameter. Functional enrichment of these networks also displayed in WikiPathways with default parameter by STRING.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCell type-specific and stimuli-dependent enhancer and DAVs analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEnhancers were identified separately across different cell types and states using DESeq2\u003csup\u003e77\u003c/sup\u003e by comparing cDNA and DNA barcode counts. Barcode counts for reference and alternative alleles were merged for each variant. Variants with an adjusted p-value \u0026lt; 0.05 and a log₂ fold change \u0026gt; 0 were classified as significant enhancers. To identify stimuli-dependent enhancers, only enhancers detected in either control or stimulation states were considered. Firstly, we compared the cDNA/DNA ratio of control versus stimulation in enhancers, normalizing by the median control ratio for each enhancer. P-values were corrected using the Benjamini–Hochberg method, and enhancers with a corrected p-value \u0026lt; 0.05 were designated as stimuli-responsive enhancers. Cell type-specific enhancers were identified using the same approach. DAVs identified across different cell types or states were overlapped; DAVs detected exclusively in one cell type were considered cell type-specific DAVs, and the same criterion was applied for different stimulation states.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCROP-seq library design and experiments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe CRISPR inhibition library (CROPi-DAVs) was designed and cloned as below. In brief, we designed sgRNAs to target 20 selected plausibly causal noncoding variants from our HepG2-MPRA screen (Supplementary Table 6). We extended 150 bp around center of each variant, then used FlashFry v1.15\u003csup\u003e89\u003c/sup\u003e and CRISPick\u003csup\u003e90\u003c/sup\u003e to detect potential sgRNAs. Finally, each variant was targeted by two sgRNAs from tops ranking by Hsu-Scott score. In addition, we included 6 non-targeting sgRNAs from the GeCKO v2 library as negative controls. All the sgRNAs sequences are shown in Supplemental Table S9.\u003c/p\u003e\n\u003cp\u003eTo create the CROPi-DAVs plasmid library, each sgRNAs oligos were annealed, and mixed together as an oligo duplexs pool. Then the\u0026nbsp;oligo\u0026nbsp;pool was cloned into BsmBI-v2 (NEB, R0739L) cut lenti-CROPseq-puro-v2\u0026nbsp;backbone. The purified ligation product was transformed into Competent-Cell Stbl3 (KTSM110L). Bacterial colonies were scraped in the following day, plasmids were isolated using the Endo-Free Plasmid Midi kit (CWBIO, CW2105S). For plasmid library QC, sgRNA sequences are amplified in a single 16-cycles PCR reaction with specially designed primers (CROP-P5i5-F and CROP-P7i7-R). Library representation was determined by Next-generation sequencing (Illumina).\u003c/p\u003e\n\u003cp\u003eThe CROPi-DAVs plasmid library\u0026nbsp;was packaged into lentivirus as below. CROPi-DAVs plasmid library was co-transfected into HEK-293T cells with the packaging plasmid mix (psPAX2 and pMD2.G). Lentivirus was collected at 48h post-transfection, following by estimating multiplicity of infection (MOI) by counting cell survival ratio upon puromycin selection (6 μg/ml) in the HepG2 cell line. For library infection, CRISPRi_HepG2 cells seeded at 30% confluency in 10-cm dish were infected with CROPi-DAVs lentivirus library at a low MOI level (~0.1) to minimize double infection. Puromycin (6 μg/ml) screening was performed at 48h post-infection and following by maintaining for another 7 days.\u003c/p\u003e\n\u003cp\u003eAfter puromycin selection and expansion, cells were captured by the 10X Chromium Controller using Chromium Next GEM Single Cell 3ʹ Reagent Kits v3.1 (10X Genomics, PN-1000268). Sample prep was performed according to protocol, holding 20~30 ng full-length cDNA per lane for sgRNA-enrichment PCR. Multiple PCR reactions were performed for sgRNA enrichment. In brief, in the first PCR reaction, full-length cDNA mix was used as template and 1st_erPCR_F and 1st_erPCR_R were used as primers. Then the first PCR product was cleaned up and amplified using the primers (2nd_erPCR_F and 2nd_erPCR_R). The final PCR was prepared in which 3rd_erPCR_F and 2nd_erPCR_R were used as primers. All PCR reactions were carried out for 10 cycles using 2x KAPA HiFi HotStart ReadyMix (KAPA Biosystems, KK2602) with annealing temperature at 62°C and 15 s extension per cycle. The whole-transcriptome scRNA-Seq libraries and the sgRNA-enrichment libraries were separately sequenced using novaXplus platform.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCROP-seq analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCell-barcode assignment in the barcode-enrichment library was performed as previously described\u003csup\u003e91\u003c/sup\u003e. In brief, to map sgRNAs alongside other genomic transcripts in individual cells, we constructed a custom reference genome by adding pseudo-genes representing sgRNA-containing 500 bp transcripts to the human genome assembly (Ensembl GRCh38 release). Each pseudo-gene included one sgRNA sequence, with 250 bp upstream of the sgRNA containing a U6 promoter and 230 bp downstream containing LTR sequences. Both the genome sequence file and transcription GTF file were modified synchronously to create the final reference genome for the Cell Ranger pipeline. The barcode-enrichment library was sequenced with a 150 bp paired-end (150PE) strategy, where sgRNA-containing sequences were located in Read 2 (R2) and cell barcodes along with UMIs were in Read 1 (R1). To meet the file format requirements of the 10X Genomics Cell Ranger pipeline, the first 98 bp of R2 and the first 28 bp of R1 were retained using cutadapt (https://doi.org/10.14806/ej.17.1.200). Then, sequences were mapped to the custom reference genome using cellranger (10x Genomics, v.7.1.0). Subsequently, we utilized the get_barcodes.py script\u003csup\u003e91\u003c/sup\u003e to assign cell barcodes to sgRNAs. Initially, a whitelist of sgRNAs used in our library was created to comply with the requirements of get_barcodes.py. The script was then executed with the parameters --all_reads --search_seq CTTGTGGAAAGGACGAAACACCG, where the search sequence corresponds to the fixed sequence preceding the sgRNA. sgRNA targets were assigned to each cell based on dominance, requiring that a sgRNA be three times more abundant than the sum of all other sgRNAs within a cell; only unique assignments were retained. For the six lanes of single-cell transcription data from CROP-seq, each lane was individually mapped to the custom reference genome using cellranger (10x Genomics, v.7.1.0), and the resulting data were aggregated into a single dataset. The aggregated single-cell RNA-seq (scRNA-seq) data were processed using Seurat (v.4.2.0)\u003csup\u003e90\u003c/sup\u003e, with doublets removed by DoubletFinder (v.2.0.4)\u003csup\u003e92\u003c/sup\u003e. High-quality cells were retained following subsetting with criteria of nFeature_RNA \u0026gt; 200, nFeature_RNA \u0026lt; 10,000, and percent.mt \u0026lt; 10. The SCT normalization method was applied, and differentially expressed genes (DEGs) between sgRNA-targeting cells and other cells were identified using the FindMarkers function with the t-test method, utilizing the SCT assay and scale.data slot. Parameters were set to logfc.threshold = 0, min.cells.feature = 1, min.cells.group = 1, and min.pct = 0 to maximize the detection of potential DEGs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTarget genes nomination analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo explore the numerous potential target genes of DAVs, we employed four distinct methods that consider linear distance, chromatin interactions, co-accessibility, and gene expression effects. Firstly, proximal genes were identified by selecting genes whose transcription start sites (TSS) were located within 20 kb of DAVs. Secondly, utilizing the previously described chromatin interaction data, we identified loop target genes by extending each loop anchor to 5 kb. Genes with promoters (TSS ±500 bp) located within one anchor and DAVs situated within the other anchor of the same loop were classified as loop target genes. Specifically, for liver promoter capture Hi-C and H3K4me3/H3K27ac Capture-C data, which provided gene information within one anchor, DAVs were required to reside within the other anchor to be considered loop targets. Thirdly, we leveraged the run_cicero and generate_ccans functions from the Signac package (v.1.8.0)\u003csup\u003e68\u003c/sup\u003e to identify co-accessible sites from scATAC-seq data. Co-accessible sites with a score greater than 0.3 were retained, and each anchor was extended to 2 kb from the center. Genes with TSS located within one anchor and DAVs within the opposing anchor of the same co-accessible site were designated as DAVs’ co-access genes. Finally, genes associated with DAVs through significant variant-gene pairs from liver eQTL data were identified as DAV eQTL genes. These four approaches collectively considered linear proximity, chromatin interactions, co-accessibility of regulatory regions, and gene expression influences to comprehensively predict target genes regulated by DAVs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInferring the gene regulatory network (GRN)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe gene regulatory network (GRN) was implemented utilizing Pando (v1.0.0)\u003csup\u003e24\u003c/sup\u003e. Liver snRNA-seq\u003csup\u003e17\u003c/sup\u003e and snATAC-seq datasets were first integrated using Signac to enable joint modeling of gene expression and chromatin accessibility. GRNs were inferred separately for MASLD and normal hepatocytes to characterize disease-associated regulatory rewiring. The infer_grn() function in Pando was applied with default parameters, using all accessible chromatin regions and expressed genes as input features. To identify transcription factor (TF)-based regulatory modules, the find_modules() function was employed to scan all accessible regions for TF-binding motifs based on the curated motif database embedded in Pando. Peaks containing TF motifs were linked to nearby genes using the “Signac” method, and the regulatory model was fitted using the “glm” method. TF modules were subsequently constructed with default settings, and only the highest-confidence interactions for each TF were retained.\u003c/p\u003e\n\u003cp\u003eVisualization of the reconstructed GRN was performed using the NetworkGraph() function, and networks were rendered with ggraph (v2.2.2.9000). In the network visualization, node color represents the log₂ fold change in gene expression between MASLD and normal hepatocytes, node size reflects PageRank centrality (a measure of network importance), and edge color encodes the statistical significance of each regulatory connection within its assigned module.\u003c/p\u003e\n\u003cp\u003eTo delineate regulatory relationships associated with disease-associated variant (DAV) target genes, a comprehensive GRN was first reconstructed using all accessible regions and genes across cell types, followed by extraction of a DAV-focused subnetwork containing only regulatory connections linked to predefined DAV target genes.\u003c/p\u003e\n\u003cp\u003eWe further profiled TF expression dynamics and motif activity within hepatocyte GRNs. TF expression levels in MASLD and normal hepatocytes were obtained using the aggregate_matrix() function in Pando. Motif activity was computed by first annotating motifs with the AddMotifs() function from Signac, followed by activity quantification using the RunChromVAR() function. Average motif activity for each TF was obtained using the wilcoxauc() function and compared between MASLD and normal states to identify differentially active TFs within the reconstructed GRN.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLuciferase reporter assay\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCRS oligos consisting of reference/alternative (ref/alt) pairs were selected based on our MPRA screen. Each 350 bp CRS sequence was synthesized as a gBlock and subsequently cloned into the pGL4.24 firefly luciferase reporter vector (Promega, E8421). Standard luciferase reporter assays were performed in HepG2, LX-2, Huh7\u0026nbsp;or\u0026nbsp;HEK 293T cells. 36 h post-transfection,\u0026nbsp;luciferase activity was quantified using the Dual-Luciferase® Reporter Assay System according to the manufacturer’s instructions. Luminescence signals were measured in a white 384-well plate using a BioTek Synergy Neo2 plate reader. The activity of firefly luciferase was normalized to Renilla luciferase activity to assess the functional activity of the tested regulatory elements.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCRIPSR deletion, inhibition and activation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor CRIPSR deletion sgRNA (Exci-sgRNA) plasmid construction, sgRNAs targeting the variants of interest were designed using CRISPOR. Exci-sgRNA oligos were annealed and ligated into the lentiCRISPR-v2-Puro vector or lentiCRISPR-v2-Blast vector, respectively. For CRIPSR inhibition (CRISPRi) and activation (CRISPRa) sgRNA plasmid construction, sgRNAs targeting the variants of interest were designed using CRISPick, the sgRNAs oligo inserts were annealed and cloned into the lenti-CROPseq-puro-v2 vector as described above. The lentivirus particles were produced, harvested and stored in aliquots at -80°C until use.\u003c/p\u003e\n\u003cp\u003eFor CRIPSR deletion, HepG2 cells were infected with Exci-sgRNA lentivirus and selected for 7 days using puromycin (6 µg/ml) and blasticidin (30 µg/ml). The deletion effect of each targeted variant was estimated by PCR amplification using primers targeting outside of the deletion sequence (Supplemental Table S9). The deleted small genomic regions were detected using agarose gel electrophoresis. For CRISPR inhibition, the CRISPRi_HepG2 cells were infected with CRISPRi-sgRNA lentivirus followed by puromycin selection (6 μg/ml) for at least 7 days. For CRISPR activation, the CRISPRa_HepG2 cells were infected with CRISPRa-sgRNA lentivirus followed by puromycin selection (6 μg/ml) and doxycycline treatment (5 µg/mL) for at least 7 days. All the sgRNAs sequences and PCR amplification primers are shown in Supplemental Table S9.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGene editing in HepG2\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePE5b system was used for SNP editing in HepG2 following the protocol as previously described\u003csup\u003e93\u003c/sup\u003e. The epegRNA and nickRNA was designed to recognize the surrounding region of the SNP using PRIDICT2.0 (\u003ca href=\"https://www.pridict.it/\"\u003ehttps://www.pridict.it/\u003c/a\u003e), then\u0026nbsp;the epegRNA dsODN for rs474513(G\u0026gt;A) and nickRNA dsODN for rs474513 were\u0026nbsp;synthetized and cloned into tevopreq1 epegRNA vector and nick-sgRNA vector respectively.\u0026nbsp;The\u0026nbsp;rs474513(G\u0026gt;A)\u0026nbsp;epegRNA plasmid,\u0026nbsp;rs474513\u0026nbsp;nick-sgRNA plasmid, pCMV-PEmax-P2A-GFP\u0026nbsp;and pEF1a-MLH1dn were co-transfected into HepG2 using Lipofectamine 3000 reagent.\u0026nbsp;48 h post-transfection, the GFP-positive cells were sorted by FACS and the subclones were identified by Sanger sequencing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRT-qPCR\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCells were collected and RNA was isolated with RNeasy Mini Kit according to the manufacturer’s instructions. For mRNA detection, RNA samples were reverse transcribed using HiScript® III RT SuperMix for qPCR (+gDNA wiper), and qRT-PCR was performed with ChamQ Universal SYBR qPCR Master Mix using Applied Biosystems QuantStudio 5 (Thermo Fisher Scientific, USA). Sequences of the qPCR primers are shown in Supplemental Table S9. Relative quantification analysis was performed using the\u0026nbsp;ΔΔCt method, with \u003cem\u003eHPRT1\u003c/em\u003e as an endogenous reference. Relative gene expression is presented as the ratio of the target gene to reference control.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHepatic-like cells differentiation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe monoclonal human iPSC lines were differentiated to hepatic-like cells as previously described\u003csup\u003e94\u003c/sup\u003e. In brief, hiPSCs were differentiated into Definitive Endoderm (DE) using STEMdiff Definitive Endoderm Kit for 4 days following the manufacturer’s instructions. For 4-9 days, cells were cultured in Advanced DMEM/F12 with 1% B27 (Cat: #A1895601), 1% KSR, 1% Glutamax, 1% MEM-NEAA, 0.5 µM A83-01, 0.25 µM Na-butyrate and 0.5% DMSO. Cells were maintained at 37°C in 5% CO\u003csub\u003e2\u003c/sub\u003e with 95% air and the medium was replaced daily. For 9-16 days, cells were cultured in Advanced DMEM/F12 with 1% B27 (Cat: #A1895601), 1% KSR, 1% Glutamax, 1% MEM-NEAA, 0.5 µM A83-01, 15 µM FH1, 15 µM FPH1, 0.1 µM Dexamethasone and 10 µM Hydrocortisone. Cells were maintained at 37°C in 5% CO\u003csub\u003e2\u003c/sub\u003e with 95% air and the medium was replaced daily. Cells were collected at Day0, Day4, Day9, Day14 and Day16, and lysed for RNA extraction and RT-qPCR detection.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHepatic organoid (HLO) differentiation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe hepatic organoid (HLO) differentiation was performed as previously described\u003csup\u003e95\u003c/sup\u003e. In brief, hiPSCs were differentiated into Definitive Endoderm (DE) using STEMdiff Definitive Endoderm Kit for 4 days following the manufacturer’s instructions. For 4–6 days, cells were cultured in Advanced DMEM/F12 with 2% B27 (Cat#17504044) and 1% N2 containing 500 ng/mL fibroblast growth factor 4 (FGF4) and 3 µM CHIR99021. Cells were maintained at 37°C in 5% CO\u003csub\u003e2\u003c/sub\u003e with 95% air and the medium was replaced daily. The FG cells were detached by Accutase and suspended with Matrigel matrix on ice. The mixture of cells and Matrigel was embedded in 30 µL drops on dishes in Advanced DMEM/F12 with 2% B27, 1% N2, 10 mM HEPES, 1% Glutamax, 1% Pen/Strep, 5 ng/mL fibroblast growth factor 2 (FGF2) , 10 ng/mL vascular endothelial growth factor (VEGF), 20 ng/mL epidermal growth factor (EGF), 3 µM CHIR99021, 0.5 µMA83-01, and 50 mg/mL ascorbic acid, and incubated in the CO\u003csub\u003e2\u003c/sub\u003e incubator for 4 days with medium changed every 2 days. The medium was then switched to Advanced DMEM/F12 with 2% B27 (Cat#17504044), 1% N2, 10 mM HEPES, 1% Pen/Strep, and 2 µM retinoic acid (RA), and incubated for further 4 days with medium changed every 2 days. The final media switch was to the hepatocyte culture medium (HCM) and the cells were incubated for another 6 days, changing the medium every 2 days.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFatty liver organoid HLOs model\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eModeling of steatohepatitis in HLOs was performed as previously reported\u003csup\u003e95\u003c/sup\u003e. HLOs was isolated from Matrigel and washed with cold PBS in 4°C for three times, then cultured with HCM media containing 5 µg/mL insulin and 300 µM oleic acid and 150 µM palmitic acid on ultra-low attachment 6 multi-well plates (Corning) to induce steatohepatitis. HLOs were collected at day 4 for immunostaining and RNA isolation. Accumulation of lipid in HLOs was measured using BODIPY 493/503.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3D Liver Spheroids\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe HepG2/LX-2 spheroids were generated following the protocol described previously\u003csup\u003e49\u003c/sup\u003e. HepG2-Exci-sgRNA stable cell lines and LX-2 cell lines were mixed in a 24:1 ratio into 96-well round-bottomed ultra-low attachment plates (Corning) at 2000 cells/well in 100 μL minimum essential medium supplemented with 10% fetal bovine serum (FBS). After 72 h, they were treated with 300 µM oleic acid and 150 µM palmitic acid for another 48h.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImmunostaining\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor HepG2 cell lines and HepG2/LX-2 3D spheroids immunostaining, cells were washed twice with PBS and fixed with 4% Paraformaldehyde (PFA) for 15min at room temperature. After this, samples were stained with 5µM BODIPY (493/503 nm) and 1 µg/ml DAPI for 15 min at 37°C.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFor LX-2 cell lines immunostaining, cells were washed twice with PBS and fixed with 4% Paraformaldehyde (PFA) for 15 min at room temperature. After this, samples were then permeated with 0.1% Triton X-100/PBS for 5 min and blocked with 10% BSA/PBS to avoid nonspecific staining for 1 h. Samples were reacted with primary antibodies diluted in 5% BSA at 4°C overnight, following by incubation with fluorophore-labeled secondary antibodies (4 μg/mL) diluted in 5% BSA/PBS at 37°C for 30 min, followed by staining DAPI (1 μg/mL) diluted in PBS at room temperature for 15 min.\u003c/p\u003e\n\u003cp\u003eFor HLOs immunostaining, HLOs were collected and washed twice with PBS and fixed with 4% PFA for 2 h at room temperature. Then, HLOs were permeabilized with 0.5% Triton X-100/PBS for 20 min at room temperature and blocked with PBS containing 0.5% Triton X-100 and 1% bovine serum albumin (BSA) for another 2 h at room temperature. HLOs were then stained with specific antibodies or 5 µM BODIPY (493/503 nm) and 1 µg/ml DAPI. After with washing with three times of DPBS containing 0.5% Triton X-100 and 0.2% BSA, HLOs were ready for imaging. All samples were imaged using confocal microscope (Nikon A1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFACS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHepG2 cell lines were detached by trypsin and wash with PBS for 3 times, following by 4% PFA fixation for 15 min at room temperature. After this, samples were stained with 5 µM BODIPY (493/503 nm) and 1µg/ml DAPI for 1.5 h at 37°C. Cells were washed in PBS and resuspended in PBS and transferred into Test Tube with Cell Strainer Snap Cap. Analysis of cells was performed using Agilent Novocyte Advanteon.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePolygenic risk score analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe PRS analysis was conducted using available data from the VA’s Million Veteran Program (MVP)\u003csup\u003e60\u003c/sup\u003e and initial MASLD-related analyses within VA Merit-funded MVP Cardiometabolic Project (MVP003/028)\u003csup\u003e4\u003c/sup\u003e. MVP is a large-scale biobank, supported by the Veterans Health Administration Office of Research and Development in the United States. The study protocols of MVP biobank and MVP Cardiometabolic Project were approved by Veterans Affairs (VA) Central Institutional Review Board (CIRB). The design, baseline demographics, and quality control procedures for MVP biobank have been previously described\u003csup\u003e60\u003c/sup\u003e. The MASLD-related phenotype and genetic analyses in MVP Cardiometabolic Projects formed the basis for the current PRS analyses\u003csup\u003e4\u003c/sup\u003e. PRS was constructed to test the MASLD predictive power of SNPs prioritized by MPRA. After removing the SNPs in linkage disequilibrium (r2 \u0026gt; 0.05), the 114 independent SNPs out of the 365 SNPs prioritized by MPRA assays were used to build PRS. A matched set of 114 SNPs, selected from meta-analyzed results of 10 different randomized draws of MPRA total SNPs, excluding 365 DAVs, served as the negative control. A set of 9 missense SNPs (PRS9; Table S8), which are the well-established MASLD causal SNPs\u003csup\u003e4\u003c/sup\u003e, served as the positive control.\u003c/p\u003e\n\u003cp\u003eFour different MASLD phenotype proxies were used to identify MASLD: MASLD-ALT, MASLD-ICD, ALT and FIB4 max progression. (1) For MASLD-ALT, MASLD cases were defined as participants with elevated ALT (\u0026gt; 40 U/L for men or \u0026gt;30 U/L for women during at least two time points at least 6 months apart within a 2-year window at any point prior to enrollment and exclusion of other causes of liver disease, chronic liver diseases or systemic conditions and/or alcohol use disorders. (2) For MASLD-ICD, MASLD cases were identified by ICD9 codes (456.2, 456.21, 571.5, 572.2 and 572.3) and ICD-10 codes (K72.9, K72.91, K74.0, K74.02, K74.1, K74.2, K74.6 and K74.69). (3) The MASLD phenotype proxy, ALT, was a continuous variable representing ALT levels. (4) Fibrosis-4 (FIB4) max progression was a surrogate marker of liver fibrosis progression by estimating FIB4 trajectory over time \u003csup\u003e96\u003c/sup\u003e. FIB4 was calculated using the formula ([age × AST]/[platelet × sqrt (ALT)]) as previously described. AST, ALT, and platelet count (in K/mm3) of all outpatient were collected from VA laboratories from January 2002 to November 2021.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo construct the PRS, logistic regression was performed between SNPs and MASLD proxies MASLD-ALT, MASLD-ICD, or FIB-4 max progression, while linear regression was used between SNPs and ALT levels. Variant weights were derived from effect sizes reported in the GWAS by Miao et al\u003csup\u003e7\u003c/sup\u003e. The regression analysis of PRS with MASLD phenotypes was performed using a linear regression model adjusted for age, sex, and principal components of genetic ancestry.\u0026nbsp;The regression coefficients from these analyses represented log-odds change in the outcome for each unit of increase in the standardized PRS.\u0026nbsp;To compare the predictive power in deciles, the individuals were divided into 10 deciles based on their PRS scores and assessed the odds ratio (OR) of having MASLD (MASLD-ALT) in each of the 10 deciles when compared to the lowest decile.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eComparison of DAV-derived PRS with published PRS models\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe polygenic risk score (PRS) constructed from disease-associated variants (DAVs) was benchmarked against the best-performing published MASLD PRS (PGS002732) derived from the MVP cohort, as well as PRS9 and PGS000655\u003csup\u003e4,27\u003c/sup\u003e. To minimize potential overfitting, all PRS models were evaluated in the \u003cem\u003eAll of Us Research Program\u0026nbsp;\u003c/em\u003e(AoU)as an independent validation cohort. AoU is a large, US-based biobank developed to improve and enable the inclusion of diverse, underrepresented populations in large-scale genetic and epidemiological research studies\u003csup\u003e97\u003c/sup\u003e. We accessed short read whole genome sequencing (srWGS) data, survey data, and electronic health record data from participants included in the v8 release of AoU. Age within AoU was calculated as participants’ age as of their consent date to AoU. Sex was determined using self-reported sex-at-birth data provided by AoU.\u003c/p\u003e\n\u003cp\u003eWe used MASLD-ALT as a proxy for MASLD, as defined above. ALT labs in units per liter were extracted using LOINC code 1742-6. Measurements less than 0 were removed, then outlier measurements were removed if they were 3 standard deviations away from the population average in log space.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDAV PRS for MASLD were calculated using the 114 LD-independent DAVs and marginal effect sizes from the Vujkovic et al . GWAS\u003csup\u003e4\u003c/sup\u003e. Published PRS were downloaded from the PGSCatalog\u003csup\u003e98\u003c/sup\u003e. All PRS were calculated using PLINK v1 with the --score sum argument. We residualized global ancestry from the PRS as previously described\u003csup\u003e99\u003c/sup\u003e, then standardized the scores. Association analyses were adjusted for age, sex, the first 10 PCs, and sequencing site. Statistical differences in effect size were estimated using a two-sided Z-test. P-values were adjusted for multiple testing per analysis using the Benjamini-Hochberg procedure.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQuantification and statistical analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData are presented as mean ± SEM. Statistical significance was determined by unpaired two-tailed student t-test, Fisher’s exact test or two-way anova analysis using GraphPad Prism software or R software; a p-value of \u0026lt; 0.05 or FDR\u0026lt;0.01 was considered significant. Specific statistical tests performed are listed in the respective figure legends or sections of the Methods.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eMatchett, K.P., Paris, J., Teichmann, S.A. \u0026amp; Henderson, N.C. Spatial genomics: mapping human steatotic liver disease. \u003cem\u003eNat Rev Gastroenterol Hepatol\u003c/em\u003e 21, 646\u0026ndash;660 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEslam, M. \u0026amp; George, J. Genetic contributions to NAFLD: leveraging shared genetics to uncover systems biology. \u003cem\u003eNat Rev Gastroenterol Hepatol\u003c/em\u003e 17, 40\u0026ndash;52 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen, Y. \u003cem\u003eet al.\u003c/em\u003e Genome-wide association meta-analysis identifies 17 loci associated with nonalcoholic fatty liver disease. \u003cem\u003eNat Genet\u003c/em\u003e 55, 1640\u0026ndash;1650 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVujkovic, M. \u003cem\u003eet al.\u003c/em\u003e A multiancestry genome-wide association study of unexplained chronic ALT elevation as a proxy for nonalcoholic fatty liver disease with histological and radiological validation. \u003cem\u003eNat Genet\u003c/em\u003e 54, 761\u0026ndash;771 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRomeo, S. \u003cem\u003eet al.\u003c/em\u003e Genetic variation in PNPLA3 confers susceptibility to nonalcoholic fatty liver disease. \u003cem\u003eNat Genet\u003c/em\u003e 40, 1461\u0026ndash;5 (2008).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAnstee, Q.M. \u003cem\u003eet al.\u003c/em\u003e Genome-wide association study of non-alcoholic fatty liver and steatohepatitis in a histologically characterised cohort(☆). \u003cem\u003eJ Hepatol\u003c/em\u003e 73, 505\u0026ndash;515 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMiao, Z. \u003cem\u003eet al.\u003c/em\u003e Identification of 90 NAFLD GWAS loci and establishment of NAFLD PRS and causal role of NAFLD in coronary artery disease. \u003cem\u003eHGG Adv\u003c/em\u003e 3, 100056 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZeng, B. \u003cem\u003eet al.\u003c/em\u003e Genetic regulation of cell type-specific chromatin accessibility shapes brain disease etiology. \u003cem\u003eScience\u003c/em\u003e 384, eadh4265 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCooper, Y.A. \u003cem\u003eet al.\u003c/em\u003e Functional regulatory variants implicate distinct transcriptional networks in dementia. \u003cem\u003eScience\u003c/em\u003e 377, eabi8654 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChiou, J. \u003cem\u003eet al.\u003c/em\u003e Single-cell chromatin accessibility identifies pancreatic islet cell type- and state-specific regulatory programs of diabetes risk. \u003cem\u003eNat Genet\u003c/em\u003e 53, 455\u0026ndash;466 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGaulton, K.J., Preissl, S. \u0026amp; Ren, B. Interpreting non-coding disease-associated human variants using single-cell epigenomics. \u003cem\u003eNat Rev Genet\u003c/em\u003e 24, 516\u0026ndash;534 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNott, A. \u003cem\u003eet al.\u003c/em\u003e Brain cell type-specific enhancer-promoter interactome maps and disease-risk association. \u003cem\u003eScience\u003c/em\u003e 366, 1134\u0026ndash;1139 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGordon, M.G. \u003cem\u003eet al.\u003c/em\u003e lentiMPRA and MPRAflow for high-throughput functional characterization of gene regulatory elements. \u003cem\u003eNat Protoc\u003c/em\u003e 15, 2387\u0026ndash;2412 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDeng, C. \u003cem\u003eet al.\u003c/em\u003e Massively parallel characterization of regulatory elements in the developing human cortex. \u003cem\u003eScience\u003c/em\u003e 384, eadh0559 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRummel, C.K. \u003cem\u003eet al.\u003c/em\u003e Massively parallel functional dissection of schizophrenia-associated noncoding genetic variants. \u003cem\u003eCell\u003c/em\u003e 186, 5165\u0026ndash;5182 e33 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSeidman, J.S. \u003cem\u003eet al.\u003c/em\u003e Niche-Specific Reprogramming of Epigenetic Landscapes Drives Myeloid Cell Diversity in Nonalcoholic Steatohepatitis. \u003cem\u003eImmunity\u003c/em\u003e 52, 1057\u0026ndash;1074 e7 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXiao, Y. \u003cem\u003eet al.\u003c/em\u003e Hepatocytes demarcated by EphB2 contribute to the progression of nonalcoholic steatohepatitis. \u003cem\u003eSci Transl Med\u003c/em\u003e 15, eadc9653 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKim, H.Y. \u003cem\u003eet al.\u003c/em\u003e Multi-modal analysis of human hepatic stellate cells identifies novel therapeutic targets for metabolic dysfunction-associated steatotic liver disease. \u003cem\u003eJ Hepatol\u003c/em\u003e 82, 882\u0026ndash;897 (2025).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGovaere, O. \u003cem\u003eet al.\u003c/em\u003e Transcriptomic profiling across the nonalcoholic fatty liver disease spectrum reveals gene signatures for steatohepatitis and fibrosis. \u003cem\u003eSci Transl Med\u003c/em\u003e 12(2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYu, F. \u003cem\u003eet al.\u003c/em\u003e Variant to function mapping at single-cell resolution through network propagation. \u003cem\u003eNat Biotechnol\u003c/em\u003e 40, 1644\u0026ndash;1653 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWiering, L., Subramanian, P. \u0026amp; Hammerich, L. Hepatic Stellate Cells: Dictating Outcome in Nonalcoholic Fatty Liver Disease. \u003cem\u003eCell Mol Gastroenterol Hepatol\u003c/em\u003e 15, 1277\u0026ndash;1292 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang, P., Kang, Q., Wu, W.S. \u0026amp; Rui, L. Hepatic Snai1 and Snai2 promote liver regeneration and suppress liver fibrosis in mice. \u003cem\u003eCell Rep\u003c/em\u003e 43, 113875 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang, Z.F., Drumea, K., Mott, S., Wang, J. \u0026amp; Rosmarin, A.G. GABP transcription factor (nuclear respiratory factor 2) is required for mitochondrial biogenesis. \u003cem\u003eMol Cell Biol\u003c/em\u003e 34, 3194\u0026ndash;201 (2014).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFleck, J.S. \u003cem\u003eet al.\u003c/em\u003e Inferring and perturbing cell fate regulomes in human brain organoids. \u003cem\u003eNature\u003c/em\u003e 621, 365\u0026ndash;372 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFairfield, C.J. \u003cem\u003eet al.\u003c/em\u003e Genome-Wide Association Study of NAFLD Using Electronic Health Records. \u003cem\u003eHepatol Commun\u003c/em\u003e 6, 297\u0026ndash;308 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGhodsian, N. \u003cem\u003eet al.\u003c/em\u003e Electronic health record-based genome-wide meta-analysis provides insights on the genetic architecture of non-alcoholic fatty liver disease. \u003cem\u003eCell Rep Med\u003c/em\u003e 2, 100437 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNamjou, B. \u003cem\u003eet al.\u003c/em\u003e GWAS and enrichment analyses of non-alcoholic fatty liver disease identify new trait-associated genes and pathways across eMERGE Network. \u003cem\u003eBMC Med\u003c/em\u003e 17, 135 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGao, C. \u003cem\u003eet al.\u003c/em\u003e Genome-wide association analysis of serum alanine and aspartate aminotransferase, and the modifying effects of BMI in 388k European individuals. \u003cem\u003eGenet Epidemiol\u003c/em\u003e 45, 664\u0026ndash;681 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEmdin, C.A. \u003cem\u003eet al.\u003c/em\u003e Association of Genetic Variation With Cirrhosis: A Multi-Trait Genome-Wide Association and Gene-Environment Interaction Study. \u003cem\u003eGastroenterology\u003c/em\u003e 160, 1620\u0026ndash;1633 e13 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEmdin, C.A. \u003cem\u003eet al.\u003c/em\u003e A missense variant in Mitochondrial Amidoxime Reducing Component 1 gene and protection against liver disease. \u003cem\u003ePLoS Genet\u003c/em\u003e 16, e1008629 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGallagher, M.D. \u0026amp; Chen-Plotkin, A.S. The Post-GWAS Era: From Association to Function. \u003cem\u003eAm J Hum Genet\u003c/em\u003e 102, 717\u0026ndash;730 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHarris, L. \u003cem\u003eet al.\u003c/em\u003e Genome-wide association testing beyond SNPs. \u003cem\u003eNat Rev Genet\u003c/em\u003e 26, 156\u0026ndash;170 (2025).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCaliskan, M. \u003cem\u003eet al.\u003c/em\u003e Genetic and Epigenetic Fine Mapping of Complex Trait Associated Loci in the Human Liver. \u003cem\u003eAm J Hum Genet\u003c/em\u003e 105, 89\u0026ndash;107 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJung, I. \u003cem\u003eet al.\u003c/em\u003e A compendium of promoter-centered long-range chromatin interactions in the human genome. \u003cem\u003eNat Genet\u003c/em\u003e 51, 1442\u0026ndash;1449 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMusso, G., Cassader, M., Bo, S., De Michieli, F. \u0026amp; Gambino, R. Sterol regulatory element-binding factor 2 (SREBF-2) predicts 7-year NAFLD incidence and severity of liver disease and lipoprotein and glucose dysmetabolism. \u003cem\u003eDiabetes\u003c/em\u003e 62, 1109\u0026ndash;20 (2013).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKang, B., Kang, B., Roh, T.Y., Seong, R.H. \u0026amp; Kim, W. The Chromatin Accessibility Landscape of Nonalcoholic Fatty Liver Disease Progression. \u003cem\u003eMol Cells\u003c/em\u003e 45, 343\u0026ndash;352 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu, D. \u003cem\u003eet al.\u003c/em\u003e Ets-1 deficiency alleviates nonalcoholic steatohepatitis via weakening TGF-beta1 signaling-mediated hepatocyte apoptosis. \u003cem\u003eCell Death Dis\u003c/em\u003e 10, 458 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMao, Z. \u003cem\u003eet al.\u003c/em\u003e ETV5 Regulates Hepatic Fatty Acid Metabolism Through PPAR Signaling Pathway. \u003cem\u003eDiabetes\u003c/em\u003e 70, 214\u0026ndash;226 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eIsaac, R. \u003cem\u003eet al.\u003c/em\u003e TM7SF3 controls TEAD1 splicing to prevent MASH-induced liver fibrosis. \u003cem\u003eCell Metab\u003c/em\u003e 36, 1030\u0026ndash;1043 e7 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBuck, M. \u0026amp; Chojkier, M. C/EBPbeta associates with caspase 8 complex proteins and modulates apoptosis in hepatic stellate cells. \u003cem\u003eJ Clin Gastroenterol\u003c/em\u003e 41 Suppl 3, S295\u0026ndash;9 (2007).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHatting, M. \u003cem\u003eet al.\u003c/em\u003e Hepatocyte caspase-8 is an essential modulator of steatohepatitis in rodents. \u003cem\u003eHepatology\u003c/em\u003e 57, 2189\u0026thinsp;\u0026ndash;\u0026thinsp;201 (2013).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang, L. \u003cem\u003eet al.\u003c/em\u003e The multiple roles of interferon regulatory factor family in health and disease. \u003cem\u003eSignal Transduct Target Ther\u003c/em\u003e 9, 282 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAbul-Husn, N.S. \u003cem\u003eet al.\u003c/em\u003e A Protein-Truncating HSD17B13 Variant and Protection from Chronic Liver Disease. \u003cem\u003eN Engl J Med\u003c/em\u003e 378, 1096\u0026ndash;1106 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCherubini, A. \u003cem\u003eet al.\u003c/em\u003e Interaction between estrogen receptor-alpha and PNPLA3 p.I148M variant drives fatty liver disease susceptibility in women. \u003cem\u003eNat Med\u003c/em\u003e 29, 2643\u0026ndash;2655 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePearson-Gallion, B. \u003cem\u003eet al.\u003c/em\u003e Abstract 321: Altered Epha2 Signaling In Non-alcoholic Fatty Liver Disease Progression. 43, A321-A321 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDatlinger, P. \u003cem\u003eet al.\u003c/em\u003e Pooled CRISPR screening with single-cell transcriptome readout. \u003cem\u003eNat Methods\u003c/em\u003e 14, 297\u0026ndash;301 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang, Y., Konrad, R.J., Ploug, M. \u0026amp; Young, S.G. APOA5 deficiency causes hypertriglyceridemia by reducing amounts of lipoprotein lipase in capillaries. \u003cem\u003eJ Lipid Res\u003c/em\u003e 65, 100578 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu, J., Afroza, H., Rader, D.J. \u0026amp; Jin, W. Angiopoietin-like protein 3 inhibits lipoprotein lipase activity through enhancing its cleavage by proprotein convertases. \u003cem\u003eJ Biol Chem\u003c/em\u003e 285, 27561\u0026ndash;70 (2010).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePennisi, G. \u003cem\u003eet al.\u003c/em\u003e ANGPTL3 Downregulation Increases Intracellular Lipids by Reducing Energy Utilization. \u003cem\u003eArterioscler Thromb Vasc Biol\u003c/em\u003e 44, 1086\u0026ndash;1097 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBergmark, B.A. \u003cem\u003eet al.\u003c/em\u003e Effect of Vupanorsen on Non-High-Density Lipoprotein Cholesterol Levels in Statin-Treated Patients With Elevated Cholesterol: TRANSLATE-TIMI 70. \u003cem\u003eCirculation\u003c/em\u003e 145, 1377\u0026ndash;1386 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNevzorova, Y.A. \u003cem\u003eet al.\u003c/em\u003e Enhanced expression of c-myc in hepatocytes promotes initiation and progression of alcoholic liver disease. \u003cem\u003eJ Hepatol\u003c/em\u003e 64, 628\u0026ndash;40 (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRess, C. \u003cem\u003eet al.\u003c/em\u003e The role of apolipoprotein A5 in non-alcoholic fatty liver disease. \u003cem\u003eGut\u003c/em\u003e 60, 985\u0026thinsp;\u0026ndash;\u0026thinsp;91 (2011).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOttensmann, L. \u003cem\u003eet al.\u003c/em\u003e Genome-wide association analysis of plasma lipidome identifies 495 genetic associations. \u003cem\u003eNat Commun\u003c/em\u003e 14, 6934 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTeratani, T. \u003cem\u003eet al.\u003c/em\u003e Lipoprotein Lipase Up-regulation in Hepatic Stellate Cells Exacerbates Liver Fibrosis in Nonalcoholic Steatohepatitis in Mice. \u003cem\u003eHepatol Commun\u003c/em\u003e 3, 1098\u0026ndash;1112 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePan, Z. \u003cem\u003eet al.\u003c/em\u003e Inhibition of MERTK reduces organ fibrosis in mouse models of fibrotic disease. \u003cem\u003eSci Transl Med\u003c/em\u003e 16, eadj0133 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMahajan, A. \u003cem\u003eet al.\u003c/em\u003e Fine-mapping type 2 diabetes loci to single-variant resolution using high-density imputation and islet-specific epigenome maps. \u003cem\u003eNat Genet\u003c/em\u003e 50, 1505\u0026ndash;1513 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBertran, L. \u003cem\u003eet al.\u003c/em\u003e The Potential Protective Role of RUNX1 in Nonalcoholic Fatty Liver Disease. \u003cem\u003eInt J Mol Sci\u003c/em\u003e 22(2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOuchi, R. \u003cem\u003eet al.\u003c/em\u003e Modeling Steatohepatitis in Humans with Pluripotent Stem Cell-Derived Organoids. \u003cem\u003eCell Metab\u003c/em\u003e 30, 374\u0026ndash;384 e6 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKachuri, L. \u003cem\u003eet al.\u003c/em\u003e Principles and methods for transferring polygenic risk scores across global populations. \u003cem\u003eNat Rev Genet\u003c/em\u003e 25, 8\u0026ndash;25 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGaziano, J.M. \u003cem\u003eet al.\u003c/em\u003e Million Veteran Program: A mega-biobank to study genetic influences on health and disease. \u003cem\u003eJ Clin Epidemiol\u003c/em\u003e 70, 214\u0026thinsp;\u0026ndash;\u0026thinsp;23 (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuang, X., Rymbekova, A., Dolgova, O., Lao, O. \u0026amp; Kuhlwilm, M. Harnessing deep learning for population genetic inference. \u003cem\u003eNat Rev Genet\u003c/em\u003e 25, 61\u0026ndash;78 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHutchinson, A., Asimit, J. \u0026amp; Wallace, C. Fine-mapping genetic associations. \u003cem\u003eHum Mol Genet\u003c/em\u003e 29, R81-R88 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLee, D. \u003cem\u003eet al.\u003c/em\u003e A method to predict the impact of regulatory variants from DNA sequence. \u003cem\u003eNat Genet\u003c/em\u003e 47, 955\u0026ndash;61 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBenaglio, P. \u003cem\u003eet al.\u003c/em\u003e Mapping genetic effects on cell type-specific chromatin accessibility and annotating complex immune trait variants using single nucleus ATAC-seq in peripheral blood. \u003cem\u003ePLoS Genet\u003c/em\u003e 19, e1010759 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHemani, G., Bowden, J. \u0026amp; Davey Smith, G. Evaluating the potential role of pleiotropy in Mendelian randomization studies. \u003cem\u003eHum Mol Genet\u003c/em\u003e 27, R195-R208 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePashos, E.E. \u003cem\u003eet al.\u003c/em\u003e Large, Diverse Population Cohorts of hiPSCs and Derived Hepatocyte-like Cells Reveal Functional Genetic Variation at Blood Lipid-Associated Loci. \u003cem\u003eCell Stem Cell\u003c/em\u003e 20, 558\u0026ndash;570 e10 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHu, W. \u003cem\u003eet al.\u003c/em\u003e Patient Adipose Stem Cell-Derived Adipocytes Reveal Genetic Variation that Predicts Antidiabetic Drug Response. \u003cem\u003eCell Stem Cell\u003c/em\u003e 24, 299\u0026ndash;308 e6 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eStuart, T., Srivastava, A., Madad, S., Lareau, C.A. \u0026amp; Satija, R. Single-cell chromatin state analysis with Signac. \u003cem\u003eNat Methods\u003c/em\u003e 18, 1333\u0026ndash;1341 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchep, A.N., Wu, B., Buenrostro, J.D. \u0026amp; Greenleaf, W.J. chromVAR: inferring transcription-factor-associated accessibility from single-cell epigenomic data. \u003cem\u003eNat Methods\u003c/em\u003e 14, 975\u0026ndash;978 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTanigawa, Y., Dyer, E.S. \u0026amp; Bejerano, G. WhichTF is functionally important in your open chromatin data? \u003cem\u003ePLoS Comput Biol\u003c/em\u003e 18, e1010378 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHao, Y. \u003cem\u003eet al.\u003c/em\u003e Integrated analysis of multimodal single-cell data. \u003cem\u003eCell\u003c/em\u003e 184, 3573\u0026ndash;3587 e29 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBulik-Sullivan, B.K. \u003cem\u003eet al.\u003c/em\u003e LD Score regression distinguishes confounding from polygenicity in genome-wide association studies. \u003cem\u003eNat Genet\u003c/em\u003e 47, 291\u0026ndash;5 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang, J., Lee, S.H., Goddard, M.E. \u0026amp; Visscher, P.M. GCTA: a tool for genome-wide complex trait analysis. \u003cem\u003eAm J Hum Genet\u003c/em\u003e 88, 76\u0026ndash;82 (2011).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGiambartolomei, C. \u003cem\u003eet al.\u003c/em\u003e Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. \u003cem\u003ePLoS Genet\u003c/em\u003e 10, e1004383 (2014).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi, H. \u0026amp; Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. \u003cem\u003eBioinformatics\u003c/em\u003e 25, 1754\u0026ndash;60 (2009).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLu, X. \u003cem\u003eet al.\u003c/em\u003e Global discovery of lupus genetic risk variant allelic enhancer activity. \u003cem\u003eNat Commun\u003c/em\u003e 12, 1611 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLove, M.I., Huber, W. \u0026amp; Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. \u003cem\u003eGenome Biol\u003c/em\u003e 15, 550 (2014).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYu, G., Wang, L.G. \u0026amp; He, Q.Y. ChIPseeker: an R/Bioconductor package for ChIP peak annotation, comparison and visualization. \u003cem\u003eBioinformatics\u003c/em\u003e 31, 2382\u0026ndash;3 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSloan, C.A. \u003cem\u003eet al.\u003c/em\u003e ENCODE data at the ENCODE portal. \u003cem\u003eNucleic Acids Res\u003c/em\u003e 44, D726-32 (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eConsortium, G.T. The GTEx Consortium atlas of genetic regulatory effects across human tissues. \u003cem\u003eScience\u003c/em\u003e 369, 1318\u0026ndash;1330 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eErnst, J. \u0026amp; Kellis, M. ChromHMM: automating chromatin-state discovery and characterization. \u003cem\u003eNat Methods\u003c/em\u003e 9, 215\u0026ndash;6 (2012).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBernstein, B.E. \u003cem\u003eet al.\u003c/em\u003e The NIH Roadmap Epigenomics Mapping Consortium. \u003cem\u003eNat Biotechnol\u003c/em\u003e 28, 1045\u0026ndash;8 (2010).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eQuinlan, A.R. BEDTools: The Swiss-Army Tool for Genome Feature Analysis. \u003cem\u003eCurr Protoc Bioinformatics\u003c/em\u003e 47, 11 12 1\u0026ndash;34 (2014).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKumar, S., Ambrosini, G. \u0026amp; Bucher, P. SNP2TFBS - a database of regulatory SNPs affecting predicted transcription factor binding site affinity. \u003cem\u003eNucleic Acids Res\u003c/em\u003e 45, D139-D144 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCoetzee, S.G., Coetzee, G.A. \u0026amp; Hazelett, D.J. motifbreakR: an R/Bioconductor package for predicting variant effects at transcription factor binding sites. \u003cem\u003eBioinformatics\u003c/em\u003e 31, 3847\u0026ndash;9 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBailey, T.L., Johnson, J., Grant, C.E. \u0026amp; Noble, W.S. The MEME Suite. \u003cem\u003eNucleic Acids Res\u003c/em\u003e 43, W39-49 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHeinz, S. \u003cem\u003eet al.\u003c/em\u003e Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. \u003cem\u003eMol Cell\u003c/em\u003e 38, 576\u0026ndash;89 (2010).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSzklarczyk, D. \u003cem\u003eet al.\u003c/em\u003e The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. \u003cem\u003eNucleic Acids Res\u003c/em\u003e 51, D638-D646 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMcKenna, A. \u0026amp; Shendure, J. FlashFry: a fast and flexible tool for large-scale CRISPR target design. \u003cem\u003eBMC Biol\u003c/em\u003e 16, 74 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDoench, J.G. \u003cem\u003eet al.\u003c/em\u003e Optimized sgRNA design to maximize activity and minimize off-target effects of CRISPR-Cas9. \u003cem\u003eNat Biotechnol\u003c/em\u003e 34, 184\u0026ndash;191 (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHill, A.J. \u003cem\u003eet al.\u003c/em\u003e On the design of CRISPR-based single-cell molecular screens. \u003cem\u003eNat Methods\u003c/em\u003e 15, 271\u0026ndash;274 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMcGinnis, C.S., Murrow, L.M. \u0026amp; Gartner, Z.J. DoubletFinder: Doublet Detection in Single-Cell RNA Sequencing Data Using Artificial Nearest Neighbors. \u003cem\u003eCell Syst\u003c/em\u003e 8, 329\u0026ndash;337 e4 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDoman, J.L., Sousa, A.A., Randolph, P.B., Chen, P.J. \u0026amp; Liu, D.R. Designing and executing prime editing experiments in mammalian cells. \u003cem\u003eNat Protoc\u003c/em\u003e 17, 2431\u0026ndash;2468 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHu, W. \u003cem\u003eet al.\u003c/em\u003e Individual-specific functional epigenomics reveals genetic determinants of adverse metabolic effects of glucocorticoids. \u003cem\u003eCell Metab\u003c/em\u003e 33, 1592\u0026ndash;1609 e7 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKimura, M. \u003cem\u003eet al.\u003c/em\u003e En masse organoid phenotyping informs metabolic-associated genetic susceptibility to NASH. \u003cem\u003eCell\u003c/em\u003e 185, 4216\u0026ndash;4232 e16 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKaplan, D.E. \u003cem\u003eet al.\u003c/em\u003e Clinical and genetic risk factors for progressive fibrosis in metabolic dysfunction-associated steatotic liver disease. \u003cem\u003eHepatol Commun\u003c/em\u003e 8(2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAll of Us Research Program Genomics, I. Genomic data in the All of Us Research Program. \u003cem\u003eNature\u003c/em\u003e 627, 340\u0026ndash;346 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLambert, S.A. \u003cem\u003eet al.\u003c/em\u003e The Polygenic Score Catalog as an open database for reproducibility and systematic evaluation. \u003cem\u003eNat Genet\u003c/em\u003e 53, 420\u0026ndash;425 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKhera, A.V. \u003cem\u003eet al.\u003c/em\u003e Whole-Genome Sequencing to Characterize Monogenic and Polygenic Contributions in Patients Hospitalized With Early-Onset Myocardial Infarction. \u003cem\u003eCirculation\u003c/em\u003e 139, 1593\u0026ndash;1602 (2019).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6984670/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6984670/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eGenome-wide association studies (GWAS) have identified nearly 100 loci associated with metabolic dysfunction-associated steatotic liver disease (MASLD), but the molecular functions of these variant alleles remain elusive, particularly when they occur in non-coding regions. Here we profiled the chromatin accessibility landscape of liver nuclei from MASLD individuals, and demonstrated these accessible genomic sites were bound by cell type-specific transcription factors (TFs) and enriched for MASLD risk variants, highlighting lineage- and disease state-specific regulation. Using a massively parallel reporter assay (MPRA), we identified hundreds of differential activity variants (DAVs) that operate in a cell type-specific manner or in a stimulus-dependent context by disrupting liver pathogenesis-associated transcriptional regulatory network. Integrative analyses combining liver eQTLs, chromatin looping, and single-cell CRISPRi screening linked these DAVs to functional target genes. Notably, we demonstrated that DAVs located near \u003cem\u003eSLC22A3\u003c/em\u003e and core regulators of triglyceride metabolism (\u003cem\u003eAPOA5\u003c/em\u003e, \u003cem\u003eANGPTL3\u003c/em\u003e, and \u003cem\u003eLPL\u003c/em\u003e) loci modulate their gene expression and contribute to altered lipid metabolism and hepatic stellate cell activation. Furthermore, these DAVs exhibit predictive power in distinguishing MASLD disease risk. Together, these multimodal integration analyses provide insights into the regulatory mechanisms of MASLD progression driven by noncoding genetic risks.\u003c/p\u003e","manuscriptTitle":"Integrative functional genomics reveals transcriptional regulatory function of risk alleles for metabolic liver disease","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-23 04:58:46","doi":"10.21203/rs.3.rs-6984670/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-genetics","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"ng","sideBox":"Learn more about [Nature Genetics](http://www.nature.com/ng/)","snPcode":"","submissionUrl":"","title":"Nature Genetics","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Research","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"ef90e7be-b5ab-4778-9518-d2c219898269","owner":[],"postedDate":"October 23rd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":56375049,"name":"Biological sciences/Genetics/Functional genomics"},{"id":56375050,"name":"Health sciences/Diseases/Metabolic disorders"}],"tags":[],"updatedAt":"2026-04-28T15:05:29+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-23 04:58:46","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6984670","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6984670","identity":"rs-6984670","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}