Alternative polyadenylation mediated the downregulation of lysophosphatidylglycerol acyltransferase 1 in metabolic dysfunction-associated steatotic liver disease

Abstract

Background

Alternative polyadenylation (APA) is a critical post-transcriptional regulatory mechanism involved in various diseases. Studies have shown dysfunction of APA-regulating factors such as SRSF10 in metabolic dysfunction-associated steatotic liver disease (MASLD). However, the downstream target genes and functional consequences remain unclear. This study investigated the role of APA in modulating LPGAT1 expression in MASLD.

Methods

Integrative analyses of bulk and single-cell RNA sequencing data from human and mouse MASLD livers were performed to identify APA changes. Functional validations were conducted using lysophosphatidylglycerol acyltransferase 1 (LPGAT1) 3’ UTR (3’ untranslated region)-knockout HepG2 cells under free fatty acid (FFA) treatment.

Results

Early hepatocyte-specific APA remodeling characterized by 3’ lengthening of metabolism-related genes, especially LPGAT1, was observed in MASLD. Despite elevated LPGAT1 mRNA levels, protein levels were suppressed in MASLD, associated with an increased usage of a proximal 3’ UTR segment enriched with miRNA binding sites. Deletion of this proximal region in HepG2 cells restored LPGAT1 protein levels and mitigated lipid accumulation under FFA exposure.

Conclusions

This study establishes a cell type-resolved APA regulatory map for the MASLD liver and identifies APA-mediated repression of LPGAT1 as a critical driver of hepatic lipid accumulation. These findings highlight APA regulation as not only a pathogenic mechanism but also a promising molecular target for therapeutic interventions aimed at combating the progression of metabolic liver disease.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12944-025-02853-7.

Background

Alternative polyadenylation (APA) is a widespread posttranscriptional mechanism that generates mRNA isoforms with variable 3’ untranslated regions (3’ UTRs), thereby modulating mRNA stability, translation efficiency, and subcellular localization [1]. APA is dynamically regulated by RNA-binding proteins (RBPs) [2] and environmental cues in a cell-type-specific manner [3], and its dysregulation has been implicated in multiple diseases, including cancer and neurodegeneration [4]. Mechanistically, lengthened 3’ UTRs promote miRNA-dependent gene expression regulation [5].

Metabolic dysfunction-associated steatotic liver disease (MASLD) is the most prevalent chronic liver condition characterized by excessive fat accumulation in the liver, which is not caused by heavy alcohol use [4]. The disease can progress into more severe forms, such as liver cirrhosis, hepatocellular carcinoma, and cardiovascular diseases [6]. MASLD is becoming an important health threat due to high-calorie diets and irregular meal patterns [7]. Although numerous transcriptomic studies have revealed altered gene expression patterns in MASLD [8–10], the contributions of posttranscriptional mechanisms, such as APA, remain underexplored. Recent work has linked dysregulated APA machinery, including the splicing factor SRSF10, to metabolic gene expression in MASLD [11]. However, a systematic understanding of cell type-specific APA events and their functional consequences in MASLD is lacking [12].

Although LPGAT1 deficiency is a known driver of hepatic steatosis, the regulatory mechanisms governing its downregulation in MASLD remain unclear. Paradoxically, recent transcriptomic data suggest that lysophosphatidylglycerol acyltransferase 1 (LPGAT1) mRNA levels may be maintained or elevated in steatotic livers, suggesting the presence of a posttranscriptional ‘brake’ that suppresses protein translation.

It was hypothesized that alternative polyadenylation (APA)-mediated 3’ UTR lengthening creates additional binding sites for negative regulators, thereby uncoupling mRNA expression from protein abundance. In this study, integrated single-cell and bulk RNA sequencing was employed to map cell type-specific APA dynamics. This approach aimed to identify specific 3’ UTR remodeling events in hepatocytes that drive the posttranscriptional repression of metabolic enzymes, specifically LPGAT1, providing a novel mechanistic explanation for its deficiency in MASLD.

Methods

In vivo animal experiments

Model

C57BL/6J mice (Charles River Laboratories, Beijing, China) (6 weeks old, approximately 19 g) were housed under controlled conditions (12 h light/dark cycle, 22–25 °C, 50–60% humidity) with free access to food and water. After one week of acclimation on a chow diet, the mice were divided into two groups: a high-fat diet (HFD) group and a control group. The diets were administered for 8 weeks [13]. All procedures were approved by the Experimental Animal Ethics Committee of Jilin University.

Blood biochemistry

The serum triglyceride (TG) (Nanjing Jiancheng, Nanjing, China), alanine aminotransferase (ALT) (Zciaibio, Shanghai, China) and low-density lipoprotein (LDL) (Nanjing Jiancheng, Nanjing, China) levels were measured enzymatically using biochemical assay kits. Serum insulin levels were determined using a mouse insulin ELISA kit (Cusabio, Wuhan, China) following the manufacturer’s protocol.

Single-cell isolation and sequencing

Mice were anesthetized with isoflurane and perfused using a two-step liver perfusion protocol [14, 15]. Liver tissues were excised, minced (to approximately 1 mm³), and enzymatically digested using a Liver Dissociation Kit (Miltenyi Biotec, California State, United States) for 15 min at 37 °C with agitation. The suspension was filtered through a 100-µm strainer, centrifuged at 30 × g for 2 min, and resuspended in RPMI 1640 supplemented with 2% BSA. The cell suspension (300–600 cells/µL) was loaded onto a Chromium Single-Cell Controller (10x Genomics, California State, United States) to generate gel beads. RNA-seq libraries were constructed using the Single Cell 3’ Library Kit V3.1 (10x Genomics, California State, United States) and sequenced on an Illumina NovaSeq 6000 with a depth of at least 100,000 reads per cell (CapitalBio Technology, Beijing, China). Single-cell RNA sequencing data that support the findings of this study have been deposited in the Gene Expression Omnibus database under the accession number GSE268613.

In vitro experiments

Cell cultures

HepG2 cells were cultured in high-glucose DMEM (Gibco, Massachusetts, USA) supplemented with 10% FBS (Gibco, Massachusetts, USA) and 1% penicillin‒streptomycin (Gibco, Massachusetts, USA) at 37 °C in a humidified 5% CO₂ atmosphere. When the cell density was 60–70%, the HepG2 cells were subjected to free fatty acid (FFA) overload by treatment with 1 mM FFA (oleic acid: palmitic acid = 2:1; oleic acid) for 48 h.

Preparation of a free fatty acid (FFA) solution

A 10 mM stock solution of oleic acid (OA) and palmitic acid (PA) at a 2:1 molar ratio was prepared by accurately weighing 0.2039 g of OA and 0.0921 g of PA. The OA was dissolved in 1000 µL of 0.01 mmol/L NaOH, and the PA was dissolved in 500 µL of the same NaOH solution. The mixtures were then heated to 70 °C and thoroughly mixed until fully dissolved. The fatty acid mixture solution was combined with 10 mL of BSA-supplemented medium, mixed thoroughly, filtered through a microporous membrane, and stored at 4 °C. For the experiments, a 1 mM FFA working solution was prepared by diluting the stock solution in culture medium at a 1:9 ratio.

Gene knockout cell line

The LPGAT1 3’ UTR (chr1:211745296–211749831) was knocked out in HepG2 cells by CRISPR ribonucleoprotein (CRISPR RNP). To increase the probability of multisite cleavage and achieve homozygosity, two sequential rounds of RNP delivery using two distinct sgRNA pairs flanking the target interval were performed. The first round yielded a heterozygous deletion; a second, independent sgRNA pair was then used to edit the remaining allele, producing a biallelic (homozygous) 3’ UTR KO. Single-cell cloning was performed to establish the final clone. All the sgRNA sequences and qPCR/PCR primers used are listed in Supplementary Table 1.

Oil red O staining

Mouse liver tissue samples were collected, washed with PBS, and frozen in OCT compound on dry ice for 10–20 s. After equilibration at − 20 °C for 15 min, 10-µm-thick frozen sections were prepared and stained with Oil Red O (Nanjing Jiancheng, Nanjing, China). HepG2 cells were fixed with 4% formaldehyde for 15 min, stained with Oil Red O for 15 min, and counterstained with hematoxylin for 5 min.

Immunohistochemistry

Liver tissues were collected and fixed in 10% formalin overnight, dehydrated in ethanol and embedded in paraffin. Five-micron-thick liver sections were prepared for histology. The primary and secondary antibodies were diluted and incubated according to the manufacturer’s instructions. Immunohistochemical staining was quantified using ImageJ software. For each sample, 3–5 nonoverlapping fields of view were captured at the same magnification and exposure settings. Images were converted to 8-bit grayscale, and the region of interest (ROI) was manually delineated to include the stained tissue area. The background signal was measured in an unstained region and subtracted from each measurement. The mean optical density (MOD) or integrated density (IOD/area) within the ROI was calculated and used as an index of staining intensity. For each sample, the average value from all the fields was taken as the final quantitative result and used for statistical analysis.

Transfection of MiRNA mimics

Chemically synthesized miRNA mimics and corresponding negative controls (mimics NC) were obtained from Comatebio (Changchun, China). Transfections were performed using Lipofectamine 3000 (Thermo Fisher Scientific, California State, United States) according to the manufacturer’s protocol. Stock solutions (100 µM) of miRNA mimics and mimics NC were diluted to achieve a final working concentration of 50 nM. Before transfection, the culture medium was replaced with 450 µL of fresh P/S-free DMEM, and 50 µL of the prepared transfection complex was gently added to each well.

Quantitative real-time PCR (qPCR)

Total RNA was extracted from HepG2 cells using the UNlQ-10 Column TRIzol Total RNA Isolation Kit (Sangon Biotech, Shanghai, China). cDNA synthesis was carried out using an AMV First Strand cDNA Synthesis Kit (Sangon Biotech, Shanghai, China). Quantitative real-time PCR (Thermo Fisher Scientific, Waltham, MA, USA) was performed with BlasTaq™ Probe 2X qPCR MasterMix (Abcam, Richmond, Canada). Relative mRNA expression levels were calculated using the 2^−ΔΔCT method and normalized to that of ACTB. The sequences of the primers used are provided in Supplementary Table 2.

Western blot

For LPGAT1 and HSP90, the protein concentrations were measured using a Bradford assay. Proteins were separated by 10% SDS‒PAGE and transferred to PVDF membranes, which were subsequently blocked with 5% skim milk for 2 h. The membranes were incubated with anti-LPGAT1 primary antibody (Abcam, Cambridge, UK) for 2 h, after which they were washed and incubated with secondary antibody in 5% skim milk for 2 h. The signal intensities were quantified, and the LPGAT1 levels were normalized to those of HSP90 (Abcam, Cambridge, UK).

For LPGAT1 and ACTB, the protein concentrations were measured using a Bradford assay. Proteins were separated by 10% SDS‒PAGE, transferred to PVDF membranes, blocked with high-efficiency Western blocking buffer (Genefist, Shanghai, China; GF1815) for 10 min at room temperature, incubated with anti-LPGAT1 primary antibody (Abcam, Cambridge, UK) and anti-ACTB (Abcam, Cambridge, UK) for 2 h, and then washed and incubated with secondary antibody for 2 h.

Dual-luciferase reporter assay

To validate the interaction between miR-219-5p and LPGAT1 3’ UTRs, a dual-luciferase reporter assay was performed. Mutant constructs were generated by introducing point mutations within the predicted miR-219a-5p seed-matching region to disrupt base pairing without altering the surrounding sequence context (wild-type: GACAATCA; mutant: CTGTTAGT). Wild-type and mutant LPGAT1 3’ UTR fragments were cloned and inserted into the pmirGLO vector and cotransfected into 293 T cells with miR-219-5p mimics or negative control using Lipofectamine 3000. Luciferase activity was measured using a dual-luciferase assay system on a SpectraMax L microplate reader.

TG induction and quantification

After 6 h of transfection, the medium was removed, and the cells were washed twice with prewarmed PBS. Each well was then incubated with FFA-supplemented medium or control medium for 48 h.

Following treatment, the intracellular TG content was determined using a Triacylglycerol Assay Kit (Nanjing Jiancheng, Nanjing, China) following the manufacturer’s instructions. The total protein concentration was measured using a BCA Protein Assay Kit (Epizyme, Shanghai, China), and the TG concentration was normalized to the protein concentration, expressed as mmol TG per g protein (mmol/gprot).

Bioinformatics analysis

scRNA-seq data analysis of liver cells

Gene count matrices were generated using Cell Ranger (v7.0.1) with the mouse reference genome GRCm38 (mm10) and analyzed with Seurat (v5.1.0) [16]. Before cells from different groups were integrated, the cells were filtered such that the number of genes was ≥ 250, the proportion of mitochondrial genes was < 20%, the total count was ≥ 500, and the log10(nFeature_RNA)/log10(nCount_RNA) was > 0.8. The R package scDblFinder (1.11.4) was used to identify potential cell doublets, and the Harmony algorithm was employed to determine batch effects [17]. Cluster analysis was performed using the first 50 principal components (PCs), and the results were visualized using UMAP. To annotate the cell clusters, prior canonical gene markers were utilized [18].

Analysis of APA events in different cell types

APA events in different cell types were identified by SCAPE [19]. This method enables the de novo identification and quantification of pA site usage in individual cells.

Identification of 3’ UTR length changes

The strand of the gene was determined, and the log2FC values of the proximal and distal pA sites were calculated.

Pathway enrichment analysis

KOBAS was used to detect enriched Reactome pathways from each set of upregulated/downregulated differentially expressed genes and APA events [20].

Bulk RNA-seq data analysis

Quality control was performed on the raw sequencing data with the following criteria: (1) removal of reads containing more than 15% ambiguous bases, (2) removal of reads shorter than 50 bp, and (3) removal of adapter sequences [21]. Alignment to the reference genome (GRCh38 or mm10) was performed using HISAT2 [22]. The 3’ UTR of LPGAT1 was divided into four segments (S1–S4) based on the positions of APA motifs and the CPM distribution across the 3’ UTR. Polyadenylation sites were identified using DaPars2, and temporal expression patterns were analyzed using Mfuzz for time series clustering [23, 24].

Correlation analysis between pA site usage and gene expression

Pearson correlation was performed between pA site usage and gene expression. A correlation was considered significant if |r| > 0.3 and P < 0.05.

Statistical analysis

Unless otherwise indicated, the data are presented as the mean ± SD (or mean ± SEM as specified in the figure legends). Comparisons between two groups were performed using Student’s t tests. For single-cell RNA-seq and APA analyses, P values were adjusted using the Bonferroni method. Differentially expressed genes were defined as those that met the following criteria: |average log₂ fold change| ≥ 0.5, detected in at least 10% of cells, and Bonferroni-adjusted P ≤ 0.05. APA events identified by SCAPE were considered significant when they were present in at least 10% of cells with a Bonferroni-adjusted P ≤ 0.05. Pearson correlation coefficients were calculated to assess the association between pA site usage and gene expression, and correlations with |r| > 0.3 and P < 0.05 were regarded as significant.

Results

Early APA remodeling in MASLD progression

To investigate the association between APA and disease progression in MASLD, liver transcriptome profiles from a cohort of 216 human liver samples (GSE135251) were analyzed, covering a spectrum of disease stages, including control, MASLD, and MASH (fibrosis stages F1–F4). PCA based on pA site usage revealed distinct separation between the control and disease groups (MASLD, MASH_F1–4), indicating that major APA shifts occur early in disease progression (Fig. 1a). Interestingly, the MASLD and MASH_F1–4groups clustered closely together, suggesting that most APA changes are established at the MASLD stage and remain relatively stable during fibrosis progression.

Fig. 1.

Early APA remodeling in MASLD progression. a Principal component analysis (PCA) of polyadenylation (pA) site usage and gene expression profiles across 216 human liver transcriptomes reveals a distinct separation between the control and disease groups (MASLD and MASH_F1-4), suggesting early APA shifts during disease progression. b Time series clustering of pA site usage and gene expression profiles using Mfuzz groups APA events into six clusters. c Correlation analysis between pA site usage and gene expression. d The expression levels of key APA regulatory factors were compared between control and MASLD liver samples. Statistical significance was determined by Student’s t test

Time series clustering using Mfuzz grouped the pA sites into six clusters (Fig. 1b). A large proportion of pA sites (47.8%, n = 5,869; Clusters P4, P5) showed sharp changes from the control to MASLD, followed by a plateau through the MASH stages (Fig. 1b). These early-shifting APA events suggest coordinated transcriptomic adaptation in response to metabolic stress prior to fibrotic remodeling. Functional enrichment analysis revealed that genes in Cluster P5 were predominantly involved in metabolic processes, indicating a potential link between APA regulation and hepatic metabolic dysfunction.

In parallel, gene expression changes across disease stages were examined (Fig. 1a and b). PCA and Mfuzz clustering revealed similar early-phase shifts, with most transcriptomic shifts also occurring during the control-to-MASLD transition. Approximately 35.8% of the genes (R3, R6; n = 3,351) displayed patterns resembling those of pA Clusters P4 and P5. Genes in RNA Cluster R6 were downregulated in patients with MASLD and remained stable across MASH stages. Pathway analysis revealed that RNA Cluster R6 genes were involved in fatty acid degradation.

Guided by the review by Mohanan et al., a core set of 16 APA regulators was curated. In the human cohorts, 10 of these 16 genes showed significant differential expression during the transition from control to MASLD (Fig. 1d), supporting the notion that APA dysregulation primarily emerges at this early disease stage [25].

To evaluate the relationship between APA and the expression of metabolic genes, the correlation between the usage of Cluster P5 pA sites and the expression of located genes was calculated, and 36.9% of P5 pA sites were correlated with gene expression levels (|r| > 0.3, P < 0.05), with 76.6% showing positive correlations and 23.4% showing negative correlations (Fig. 1c).

These results suggest early involvement and a regulatory role of APA in metabolic genes in the progression of MASLD.

Liver APA Altas in MASLD

To explore the cell-specific alterations in APA events in MASLD, single-cell RNA sequencing was performed on liver cells from chow-fed (n = 4) and HFD-fed mice (n = 4) after 8 weeks of diet intervention. Compared with chow-fed controls, HFD-fed mice exhibited increased body weight and significantly elevated serum TG, LDL-C, insulin, and ALT levels, supporting the presence of dyslipidemia/insulin resistance and hepatocellular injury [26]. Hepatic lipid accumulation was further confirmed by pronounced Oil Red O staining in frozen liver sections from the HFD group (Fig. 2a and b). In summary, HFD and chow groups can be utilized to investigate cell-specific APA changes in MASLD.

Fig. 2.

Single-cell alternative polyadenylation (APA) landscape of MASLD mouse liver. a Oil Red O staining shows increased hepatic lipid accumulation in HFD-fed mice compared with chow-fed controls. Scale bar: 50 μm. b Body weight gain and serum levels of ALT, insulin, and LDL after 8 weeks of feeding. P< 0.05, Student’s t test. c Major liver cell types identified by scRNA-seq, shown via UMAP, cell counts, and marker gene expression. d Numbers of significantly increased and decreased pA site usage per cell type and Reactome pathway enrichment of APA-regulated genes with 3' UTR lengthening or shortening in LSECs, HEPs, and Kupffer cells. e Differentially expressed genes were identified in each cell type, and Reactome pathway enrichment was performed for up- and downregulated genes

scRNA-seq analysis was performed on liver cells from HFD- and chow-fed mice. Overall, 19,036 high-quality liver cell transcriptomes were collected and analyzed, including 8,833 from the chow group and 10,203 from the HFD group. Twelve major liver cell types were identified, with hepatocytes (HEPs), liver sinusoidal endothelial cells (LSECs), and Kupffer cells being the dominant populations (Fig. 2c).

Differential gene expression analysis was performed on three major cell populations from single-cell RNA-seq data: LSECs, HEPs and Kupffer cells. The enrichment analysis revealed the following results: HEPs and LSECs: the upregulated genes were primarily associated with fatty acid metabolism pathways, supporting a role for these cells in lipid accumulation and related processes during MASLD; Kupffer cells: the upregulated genes were enriched in apoptosis, NF-kappa B signaling, and leukocyte transendothelial migration, suggesting a role for these cells in immune response activation and macrophage recruitment during MASLD (Fig. 2e). These findings confirm that the mouse model exhibits the characteristic features of MASLD [27].

A total of 3352 significantly different pA sites were detected (Supplementary Table 3) across the primary cell types (adjusted P value < 0.05), predominantly in LSECs, HEPs and Kupffer cells (Fig. 2d). Classification of APA-regulated genes revealed enrichment of 3’ UTR-lengthened transcripts in metabolism and chylomicron remodeling pathways in hepatocytes, VEGF signaling in LSECs, and immune response pathways in Kupffer cells.

These results establish a comprehensive APA atlas in MASLD mouse livers, highlighting hepatocyte-specific 3’ UTR lengthening of metabolism- and lipid transportation-related genes as a prominent feature of MASLD pathogenesis [28].

APA leads to a deficiency of LPGAT1 in MASLD

As an enrichment of 3’ UTR-lengthened metabolic genes was detected in hepatocytes in HFD-fed mice, the differential expression levels of the genes with longer 3’ UTRs in the hepatocytes were investigated. The results revealed consistent downregulation of pA site usage in metabolism-related genes across human and mouse livers, with LPGAT1, FDX1, and SLC27A2 showing the most conserved trends (Fig. 3a). As LPGAT1 had the longest 3’ UTR, it was selected for further investigation.

Fig. 3.

APA leads to a deficiency of LPGAT1 in MASLD. a Overlap of APA-related genes in metabolic pathways between human bulk liver RNA-seq and mouse hepatocytes (HEPs) from single-cell RNA-seq and alteration of proximal pA site usage in metabolism-related genes. b-c mRNA expression levels of Lpgat1. d mRNA expression levels of Acaca,Hspa5, and Xbp1 in HEPs from HFD- and chow-fed mice. P < 0.05, Student’s t test. e Immunohistochemical staining of Lpgat1 protein in liver sections from MASLD and chow-fed mice. Scale bar: 50 μm. Quantification of LPGAT1 IHC staining intensity from (e). Data are presented as the mean ± SD (n = 5/group). P < 0.05, Student’s t test. f Lpgat1 protein levels in the livers of HFD-fed mice were measured by Western blotting, with HSP90 as a loading control. Quantification ofLPGAT1 protein levels relative to those of ACTB. Data are presented as the mean ± SD from five independent experiments. P values were determined using Student’s t test (P < 0.05 *, P < 0.01 **, P < 0.001 ***)

Lpgat1 is an acyltransferase that catalyzes the remodeling of phosphatidylglycerol (PG), a mitochondrial phospholipid implicated in various metabolic diseases. Multiple studies have reported that Lpgat1 deficiency contributes to MASLD. Previous studies have shown that knockout of Lpgat1 in mice results in characteristics of MASLD, with genes related to lipid biosynthesis (Acaca) and the endoplasmic reticulum (Hspa5, Xbp1) being upregulated [29]. Therefore, the mRNA expression levels of these genes in hepatocytes were validated by single-cell sequencing. The results revealed that the expression levels of Acaca, Hspa5 and Xbp1 were upregulated; however, in contrast to the findings of previous reports, Lpgat1 mRNA expression was also upregulated (Fig. 3b-d). The LPGAT1 mRNA levels across the human cohorts were also quantified. Along the MASLD trajectory, LPGAT1 expression increased from the control stage to MASLD, followed by a decrease at more advanced stages. Specifically, compared with the control, LPGAT1 expression tended to increase in MASLD, which is consistent with the mouse data (Supplementary Fig. 3a and b).

Finally, the protein level of Lpgat1 was evaluated through Western blotting and immunohistochemical staining. Despite the upregulated mRNA level, the protein level of Lpgat1 was significantly reduced (Fig. 3e and f). Taken together, these results indicated that a HFD may induce the alternative usage of Lpgat1 3’ UTRs through APA, further resulting in a significant reduction in Lpgat1 protein levels.

Increased usage of LPGAT1 proximal to 3’ UTRs is strongly associated with MASLD

APA can considerably affect posttranscriptional gene regulation by altering the length of 3’ UTRs. To accurately characterize the alteration in Lpgat1 3’ UTR length in MASLD, the expression levels of Lpgat1 3’ UTRs in the RNA-seq dataset of the bulk liver tissue of MASLD mice and humans were quantified (Fig. 4a and b). The results indicated consistent 3’ UTR lengthening of LPGAT1 in MASLD mouse and human livers. The expression of a segment near the last exon of Lpgat1 (chr1:191778535:191782596, mm10; chr1:211743457:211749898, GRCh38) increased, indicating increased usage of 3’ UTRs in this segment compared with that in chow-fed mice. Intriguingly, a common APA motif, “AAUAAA”, was found to align well with the segment with increased usage.

Fig. 4.

Increased usage of LPGAT1 proximal 3' UTRs is associated with MASLD. a Expression of the LPGAT13' UTR in MASLD patient bulk tissue RNA-seq data, with the 3' UTR divided into four segments (S1–S4) based on APA motif distribution and CPM. b Expression of the LPGAT1 3' UTR in HEPs from mice and humans, with the 3' UTR divided into 100 bp bins and counts per million (CPM) calculated. The APA motif (AAUAAA) is marked by a dashed line. c Quantification of LPGAT1 3' UTR (S1–4) expression in HepG2 cells (S1–4) and mice (region 1 and region 2), as measured by qPCR. P < 0.05 *,P < 0.01 **, P < 0.001 ***, Student’s t test. d Predicted miRNA binding sites in the human and mouse LPGAT1 3' UTRs determined using TargetScan, highlighting conserved miRNAs targeting the proximal 3' UTR. The yellow border represents conserved miRNA families, and the colored boxes represent the binding probabilities of the miRNAs (8mer > 7mer-m8 > 7mer-A1 > 6mer). e 293T cells were cotransfected with wild-type (LPGAT1-WT) or mutant (LPGAT1-MUT) 3' UTR reporter constructs and either miR-219a-5p mimics or a negative control (NC). Statistical significance was determined by Student’s t test (P < 0.05 *, P < 0.01 **, P < 0.001 ***)

Whether the differential usage of pA sites in Lpgat1 is directly induced by a HFD was further examined using the human hepatocyte cell line HepG2. Compared with control cells, FFA-treated cells displayed dysfunction in lipid metabolism (Supplementary Fig. 2a). The expression of Lpgat1 3’ UTRs by segment was quantified using qPCR. The results revealed that the expression levels of the proximal 3’ UTRs (S2–4) were significantly greater in FFA-treated cells than in control cells (Fig. 4c). In HFD-fed mice, the LPGAT1 3’ UTR was quantitatively profiled by separating the proximal segment (region 1) from the distal segment (region 2). Consistent with the results in human samples, the proximal 3’ UTR (region 1) was significantly upregulated in the HFD-fed group, whereas the distal 3’ UTR (region 2) was not significantly altered (Fig. 4c). These in vivo data mirror the cell culture findings and indicate a preferential increase in the proximal 3’ UTR under steatotic stress. These findings indicate that the usage of LPGAT1 proximal to 3’ UTRs is increased by HFD or FFA treatment. This increased usage was consistent across the stages of MASLD.

The potential regulatory mechanisms underlying the increased usage of LPGAT1 proximal to 3’ UTRs and MASLD were further investigated. Because miRNAs play key roles in the posttranscriptional regulation of mRNAs, the miRNA binding sites in the proximal or distal fragments of the LPGAT1 3’ UTR for both humans and mice were predicted, and the binding sites of multiple conserved miRNAs in the proximal 3’ UTRs were identified. Of these, miR-219-5p, miR-29-3p, miR-129-3p, and miR-135-5p were exclusive to this proximal region. Notably, the binding sites of miR-219-5p were conserved across humans and mice (Fig. 4d). To clarify the interaction between the proximal 3’ UTR and miR-219a-5p, a dual-luciferase reporter assay was performed, which revealed that miR-219a-5p can bind to the proximal 3’ UTR of LPGAT1 (Fig. 4e).

Impact of LPGAT1 proximal 3’ UTRs on hepatocyte lipid metabolism

To further explore the impact of LPGAT1 proximal 3’ UTRs (pUTRs) on hepatocyte lipid metabolism, LPGAT1-pUTR knockout (KO) HepG2 cells and a matching wild-type (WT) cell line were generated. Their genotypes were subsequently confirmed using PCR and Sanger sequencing (Supplementary Fig. 1a and b).

WT and KO HepG2 cells were cultured with 1 mM FFA-supplemented medium or DMEM (as a control). Compared with their respective controls, both WT and KO cells accumulated more lipid droplets upon FFA treatment, but compared with WT-FFA cells, KO-FFA cells exhibited milder lipid accumulation (Fig. 5a). In addition, FFA treatment increased the intracellular triglyceride (TG) contents in both KO and WT cells. However, compared with the FFA-WT cells, the FFA-KO cells presented markedly lower TG levels (Fig. 5b). Western blot analysis revealed that LPGAT1 protein levels in the KO-FFA cells were comparable to those in the WT-control and KO-control groups but significantly higher than those in the WT-FFA cells (Fig. 5c), indicating that LPGAT1-pUTR knockout alleviated the FFA-induced suppression of LPGAT1 expression in MASLD.

Fig. 5.

Knockout of LPGAT1 proximal 3' UTRs protected HepG2 cells from steatosis. a Oil Red O staining of WT and KO HepG2 cells under control and FFA treatment conditions. Red staining indicates lipid droplets. Scale bars, 50 μm. b Triglyceride (TG) accumulation in WT and KO HepG2 cells under control and FFA treatment conditions. The data are presented as the mean ± SEM from 3 independent experiments. c LPGAT1 protein levels in cell models measured by Western blot, with HSP90 as a loading control. Quantification ofLPGAT1 protein levels relative to those of HSP90. d TG quantification of WT and LPGAT1 3’ UTR-KO HepG2 cells transfected with miR-219a-5p mimics or NC under DMEM or FFA treatment (n = 3). e Western blot of LPGAT1 protein with ACTB as a loading control. The data are presented as the mean ± SD from three independent experiments. Adjusted P values were obtained using Student’s t test followed by Benjamini–Hochberg correction for multiple testing (adjusted P < 0.05*, adjusted P < 0.01 **, adjusted P < 0.001 ***)

Collectively, these results indicated that the knockout of the pUTR of LPGAT1 mitigates the lipid accumulation induced by excessive FFAs through the restoration of LPGAT1 protein levels.

The miR-219a-5p mimics construct was transfected into wild-type HepG2 cells and a CRISPR line lacking the proximal LPGAT1 3’ UTR (pUTR-KO). The cells were subsequently cultured in DMEM or with free fatty acids (FFAs), generating six conditions (WT + mimics + DMEM, WT + mimics + FFA, WT + NC + DMEM, WT + NC + FFA, KO + mimics + FFA, and KO + NC + FFA) (Fig. 5d and e).

WT + mimics + DMEM vs. WT + NC + DMEM. This comparison was not statistically significant and explained the small upward shift as baseline variability in the short-3’ UTR (site-absent) context. These findings support the notion that mir-219a-5p has little effect on the post-transcriptional regulation of Lpgat1 because of the low expression of the 3’ UTR under investigation in WT cells under normal conditions (DMEM), which is consistent with the limited functional engagement of proximal 3’ UTR sites (Fig. 5d).

KO + NC + FFA vs. KO + mimics + FFA (3’ UTR-KO + FFA). This contrast was statistically significant. Although the miR-219a-5p binding site in the distal 3’ UTR was absent in KO cells, mir-219a-5p mimics decreased the protein expression of Lpgat1. This finding indicated a marginal regulatory effect of this microRNA, which is independent of the 3’ UTR under investigation.

KO + NC + DMEM and KO + mimics + DMEM were added. The protein levels of the included groups were largely in the order of WT + DMEM > KO + FFA > WT + FFA. The increase in the number of genes in the KO + FFA group compared with that in the WT + FFA group treated with NC indicated that other regulatory mechanisms, e.g., other micro-RNAs, might also rely on this 3’ UTR to function. However, the levels in the WT + DMEM groups were greater than those in the FFA-treated groups, which indicates the complexity of the mechanisms induced by FFA. The added KO baseline lanes are shown in Supplementary Fig. 5.

With respect to FFAs, triglyceride levels were significantly lower in the KO + mimics + FFA group than in the WT + mimics + FFA group, indicating that deleting the proximal 3’ UTR attenuated the miR-219a-5p-dependent steatotic response. Similarly, the abundance of LPGAT1 protein was markedly greater in the KO + mimics + FFA group than in the WT + mimics + FFA group, demonstrating the alleviation of miR-219a-5p-mediated repression (Fig. 5e).

Together, these data support a model in which miR-219a-5p drives lipid accumulation by suppressing LPGAT1 protein through binding sites within the proximal 3’ UTR that become functionally engaged in the FFA condition.

Discussion

Recent findings have indicated that LPGAT1 is a key lipid-sensitive enzyme that influences lipid metabolism and energy balance [30]. Consistently, LPGAT1-knockout mice are protected from HFD-induced obesity but develop hepatic steatosis and insulin resistance, indicating a crucial role for LPGAT1 in the partitioning of dietary lipids between liver and adipose tissue [29]. Furthermore, miR-30c has been suggested to be a potential posttranscriptional regulator of LPGAT1, although its direct binding and regulatory activity have not been confirmed [31]. Although genetic variants in LPGAT1 have been linked to obesity and lipid metabolism disorders, the regulatory mechanisms governing its expression, particularly in steatotic liver disease, remain poorly understood [32]. This study addresses this gap by revealing a posttranscriptional mechanism whereby alternative polyadenylation (APA) reshapes LPGAT1 mRNA isoforms, promoting susceptibility to miRNA-mediated repression in MASLD.

This study revealed early and cell type-specific APA remodeling, with hepatocytes displaying prominent 3’ UTR lengthening in metabolism-related genes. Among these, LPGAT1 exhibited a pronounced APA shift toward proximal polyadenylation site usage, leading to an expanded repertoire of miRNA binding sites, including that of miR-219-5p. Functional experiments confirmed that excising the proximal 3’ UTR restored LPGAT1 protein expression and ameliorated lipid accumulation in steatotic hepatocytes. These findings reveal a direct molecular mechanism connecting APA regulation with hepatocellular lipid homeostasis in MASLD. The findings in this study highlight the functional consequences of APA in modulating LPGAT1 expression and hepatic lipid accumulation; however, the upstream triggers of APA remodeling remain elusive. Previous work has implicated dysregulated RNA binding proteins (RBPs), such as SRSF10, in global APA shifts in metabolic diseases [11]. Whether nutrient-sensitive pathways (e.g., mTOR and insulin signaling) modulate the APA machinery under HFD conditions or whether inflammation-driven signaling contributes to APA remodeling in steatotic livers warrants further investigation.

Overall, this study contributes to the understanding of posttranscriptional regulation in MASLD and highlights APA as an early and targetable event in disease progression. Targeting APA regulators or specific miRNAs, such as miR-219-5p, may offer novel therapeutic avenues to restore metabolic gene expression and alleviate hepatic steatosis.

Strengths and limitations

The primary strength of this study lies in the integrative analysis of bulk and single-cell RNA sequencing data from both human and mouse cohorts, which enabled the construction of a high-resolution, cell-type-specific APA regulatory map for MASLD. Unlike previous studies that focused broadly on transcriptomic changes, this work specifically pinpointed the 3’ UTR lengthening of LPGAT1 in hepatocytes as a key posttranscriptional event. Furthermore, the functional consequences of this APA event were mechanistically validated using CRISPR/Cas9-mediated 3’ UTR editing and specific miRNA mimics, providing a direct causal link between 3’ UTR remodeling, protein downregulation, and lipid accumulation. This multilayered approach effectively bridges the gap between bioinformatics prediction and molecular function.

Despite these insights, several limitations should be acknowledged. First, manipulation of the LPGAT1 3’ UTR or miR-219a-5p was not performed in vivo. Although existing knockout models have defined the enzymatic role of LPGAT1, definitive validation of the APA-mediated regulatory mechanism in the liver environment requires sophisticated tools, such as hepatocyte-targeted AAV-mediated gene editing, which represents an important direction for future research. Second, lipidomic profiling and direct mitochondrial functional assays were not conducted in the LPGAT1 3’ UTR-KO model. Although the data demonstrate that restoring LPGAT1 protein levels alleviates triglyceride accumulation, the specific effects on phospholipid composition and mitochondrial respiration remain to be fully elucidated. Finally, although SRSF10 has been implicated in metabolic APA shifts, the precise upstream signaling pathways triggered by high-fat diet feeding that orchestrate this specific APA remodeling warrant further investigation.

Conclusions

This study establishes a cell-type-resolved APA regulatory map of the MASLD liver, identifying the extensive 3’ UTR lengthening of metabolic genes, particularly LPGAT1, as a critical driver of hepatic steatosis. The data demonstrate that this specific posttranscriptional remodeling uncouples LPGAT1 mRNA expression from protein abundance by creating an “miRNA trap” involving miR-219-5p, thereby suppressing lipid remodeling capacity.

These findings highlight APA dysregulation as a novel molecular pathology that complements traditional transcriptional profiling. Clinically, these findings suggest that therapeutic strategies specifically designed to block the interaction between miR-219-5p and the proximal 3’ UTR of LPGAT1—for instance, the use of steric-blocking antisense oligonucleotides (ASOs) or hepatocyte-targeted miRNA inhibitors—could serve as precision medicine approaches to restore hepatic lipid homeostasis. Furthermore, recognizing APA shifts as early molecular events offers a new avenue for stratifying patients at risk before irreversible damage occurs, ultimately aiming to improve long-term health outcomes in the growing MASLD population.

Abbreviations

APA

Alternative polyadenylation

MASLD

Metabolic dysfunction-associated steatotic liver disease

MASH

Metabolic dysfunction-associated steatohepatitis

UTR

Untranslated region

pUTR

Proximal untranslated region

pA site

Polyadenylation site

LPGAT1

Lysophosphatidylglycerol acyltransferase 1

RBP

RNA-binding protein

miRNA

microRNA

scRNA-seq

Single-cell RNA sequencing

RNA-seq

RNA sequencing

HFD

High-fat diet

HEP

Hepatocyte

LSEC

Liver sinusoidal endothelial cell

qPCR

Quantitative real-time PCR

FFA

Free fatty acid

TG

Triglyceride

ALT

Alanine aminotransferase

LDL

Low-density lipoprotein

WT

Wild type

KO

Knockout

IHC

Immunohistochemistry

FISH

Fluorescence in situ hybridization

PC

Principal component

PCA

Principal component analysis

CRISPR

Clustered regularly interspaced short palindromic repeats

sgRNA

Single-guide RNA

RNP

Ribonucleoprotein

Authors’ contributions

Wei Feng designed the project and bioinformatics analysis and wrote the manuscript. Yuxin Liu performed cell biological and histopathological analyses. Yunxiao Zhang, Aowen Tian, Miaoran Zhang and Peng Xu contributed to the imaging analysis and data curation. Chang Shu, Jianping Wen, Jianli Yang, Baiyu Qi, Wenjin Qiu and Zhengwen An helped write, review and edit the manuscript. Peng Chen was responsible for the overall content of this study.

Funding

This work was supported by the Natural Science Foundation of Jilin Province, China (20240101272JC), to P.C. This study was also supported by the National Key R&D Program of China (No. 2022YFC2504200), the National Natural Science Foundation of China (No. 82270960), and the Science & Technology Development Talent Project of Jilin Financial Department, Jilin, China (No. JCSZ2021893-35), to Z.A.

Data availability

The single-cell data can be accessed in the Gene Expression Omnibus database under GSE268613. The analysis code is available at https://github.com/fatcat-del/3UTR.

Declarations

Ethics approval and consent to participate

All animal experiments were conducted in accordance with institutional guidelines and were approved by the Experimental Animal Ethics Committee of Jilin University. The human transcriptomic datasets analyzed in this study were obtained from public databases (mice: GSE239747; human: GSE135251), and appropriate informed consent and ethical approval were obtained from all original studies.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.