A Tm6sf2^167K KI mouse model was created using CRISPR/Cas9 technology on the C57BL/6J background (Nanjing Institute of Biotechnology, Nanjing, China). An sgRNA targeting the mouse Tm6sf2 gene was constructed using the Cas9/RNA system gene-targeting technique (sgRNA is detailed in Supplementary Table 1). Simultaneously, a donor vector carrying the Tm6sf2-E167K fragment was constructed and co-injected with the Cas9 system into mouse zygotes. Sequencing was performed to validate the genotypes, and primer information is shown in Supplementary Table 2. To generate a mouse model of MASLD, 8-weekold male mice from both the Tm6sf2^167K/K (mutant) homozygous group and the Tm6sf2^167E/E (wild type, WT) control group were randomly divided into two groups, and fed either a control diet (CD) or a HFD (60% fat; Research Diets, New Brunswick, NJ, USA) (n=6 per group) [16]. The Tm6sf2^167E/E and Tm6sf2^167K/K mice were created by mating Tm6sf2167E/K heterozygous mice. To perform the experiment with PC containing C18:3 (PC18:3n-3, ZL-159048, STANDARDS, Shanghai, China), 8-week-old male Tm6sf2^167E/E and Tm6sf2^167K/K mice were subjected to dietary intervention using PC containing C18:3. PC containing C18:3 was dissolved using soybean oil (S110245; Aladdin, Shanghai, China) at a concentration of 25 mg/mL. The regular (once every 3 days) and quantitative (90 mg/kg PC containing C18:3) gavage was performed for 16weeks. Soybean oil served as the control reagent (n=5-6 per group) [27]. All mice were housed under constant temperature (22°C) and humidity (40–60%) conditions with a light/dark cycle of 12 hours. Sixteen weeks later, the mice were fasted for 6 hours and then euthanized. All mice were simultaneously subjected to experimental treatments to minimize potential confounders. All animal experiments were approved by the Medical Animal Protection Committee of Qingdao University.

293T cell line was purchased from Procell (Wuhan, China). 293T cells were cultured in high-glucose DMEM medium containing 10% fetal bovine serum (Biological Industries, Beit-Haemek, Israel). Mouse primary hepatocytes were obtained from the 8-week-old male Tm6sf2^167E/E and Tm6sf2^167K/K mice. The extraction steps for mouse primary hepatocytes could be found in the previous studies [28,29]. Mouse primary hepatocytes were cultured in high-glucose DMEM medium containing 2% fetal bovine serum, 100 nM dexamethasone (Solarbio, Beijing, China), and 100 nM insulin (Solarbio). Cells were incubated at 37°C in a humidified incubator with 5% CO₂. Cells were treated with 0.5 mM free fatty acids (FFAs, oleic acid: palmitic acid=2:1) for 24 hours to induce cellular lipid degeneration. Oleic acid and palmitic acid were obtained from Sigma-Aldrich (St. Louis, MO, USA).

Peripheral blood samples of 143 patients, who met the diagnostic criteria for MASLD [1], were collected in this study at Qingdao Municipal Hospital. Blood cells were used to detect the TM6SF2 rs58542926 genotype, with details about the primers provided in Table S3. A total of five healthy controls (HC) and 10 MASLD patients with the TM6SF2 E167K variant were identified. Additionally, six HC and 11 MASLD patients without the TM6SF2 E167K variant were randomly selected as controls. Twenty-one MASLD patients underwent magnetic resonance imaging (MRI) proton density fat fraction (PDFF) examination to assess hepatic fat content. The basic information and clinical characteristics of the participants are outlined in Table S4. Informed consent was obtained from all the participants. The entire study was conducted in accordance with the Declarations of Helsinki and Istanbul. This study was approved by the Ethics Committee of Qingdao Municipal Hospital.

Quantitative data were presented as mean and standard deviation (SD). Differences between the two groups were evaluated using the two-tailed Student’s t test or Mann– Whitney U test. Pearson correlation analysis was used to assess the correlation between two parameters. The receiver operator characteristic (ROC) curve analysis was used to determine the area under the curves (AUC) value, sensitivity, and specificity. Statistical analysis was performed using GraphPad Prism 9.0 software (GraphPad Software, San Diego, CA, USA). P-values lower than 0.05 were considered statistically significant.

Additional Materials and Methods were described in Supplementary information.

To investigate the role of the TM6SF2 E167K variant during MASLD progression, we utilized CRISPR/Cas9 technology to change the 499th nucleotide of the mouse Tm6sf2 gene coding sequence from G to A and substitute lysine (K) for glutamate (E) (Fig. 1A). Genetic sequencing confirmed the point mutation at the 499th nucleotide of the Tm6sf2 gene in Tm6sf2^167K KI mice (Fig. 1B). The Tm6sf2^167E/E and Tm6sf2^167K/K mice were fed with HFD for 16 weeks to establish the MASLD model (Fig. 1C). There were no significant changes in the Tm6sf2 mRNA or TM6SF2 protein levels between the Tm6sf2^167E/E and Tm6sf2^167K/K mice fed with either CD or HFD (Fig. 1D and E). A 16-week HFD resulted in significant increases in body weight, liver weight, and hepatic index in the Tm6sf2^167K/K mice compared with the Tm6sf2^167E/E mice (Fig. 1F–I). Additionally, hematoxylin and eosin (H&E) staining and Oil Red O staining showed that the Tm6sf2^167K/K mice displayed a more severe degree of steatosis compared with the Tm6sf2^167E/E mice under HFD (Fig. 1J). The TG content in liver tissue was higher in the Tm6sf2^167K/K mice than that in the Tm6sf2^167E/E mice fed with HFD (Fig. 1K). In contrast, the plasma TG and total cholesterol (TC) levels showed no significant differences between the Tm6sf2^167E/E and Tm6sf2^167K/K mice (Supplementary Fig. 1). Then primary hepatocytes were isolated from the Tm6sf2^167E/E and Tm6sf2^167K/K mice and treated with FFAs. The results of both Oil Red O staining and TG detection showed that primary hepatocytes from the Tm6sf2^167K/K mice displayed a more significant accumulation of lipids compared with the Tm6sf2^167E/E mice (Supplementary Fig. 2). Taken together, these results indicate that the TM6SF2 E167K variant significantly increases the sensitivity of mouse liver to HFD, making it more prone to developing hepatic steatosis. Moreover, the MASLD activity score in the Tm6sf2^167K/K mice was higher than that in the Tm6sf2^167E/E mice under HFD (Fig. 1L). Simultaneously, the Tm6sf2^167K/K mice fed with HFD had significantly elevated plasma alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels compared with the Tm6sf2^167E/E mice (Fig. 1M and N), suggesting that the TM6SF2 E167K variant could also exacerbate hepatic injury under HFD stimulation.

To further investigate the impact of the TM6SF2 E167K variant on hepatic lipids composition, we used liquid chromatography-mass spectrometry (LC-MS) to conduct untargeted lipidomics analysis of liver tissue from the Tm6sf2^167E/E and Tm6sf2^167K/K mice fed with CD or HFD. The results revealed that the quality-control (QC) samples clustered together and were located in the middle of each group, indicating good experimental reproducibility and distinct separation among the groups (Fig. 2A). A total of 27 lipid subclasses and 1231 lipid molecules were identified in the lipidomics analysis (Fig. 2B). Quantitative analysis of the identified lipids subtypes revealed that TG and PC were the two most predominant subtypes in terms of content (Supplementary Fig. 3). Hepatic TG content was significantly higher in the Tm6sf2^167K/K mice than in the Tm6sf2^167E/E mice under HFD (Fig. 2C). Under both the CD and HFD conditions, hepatic PC content was significantly lower in the Tm6sf2^167K/K mice compared with the Tm6sf2^167E/E mice (Fig. 2D). To further analyze the impact of the TM6SF2 E167K variant on the hepatic contents of unsaturated TG and PC, lipids were categorized into three groups, namely saturated fatty acid (SFA)-lipids, monounsaturated fatty acid (MUFA)-lipids, and polyunsaturated fatty acid (PUFA)-lipids. Under HFD conditions, compared with the Tm6sf2^167E/E mice, the Tm6sf2^167K/K mice showed a significant increase in the hepatic contents of monounsaturated TG and polyunsaturated TG, whereas the level of polyunsaturated PC was noticeably reduced (Fig. 2E and F). Furthermore, by analyzing the ratio of PC to TG in different groups, we also found that the HFD-induced Tm6sf2^167K/K mice displayed a significant decrease in the ratio of polyunsaturated PC to TG compared with the Tm6sf2^167E/E mice (Fig. 2G). Additionally, under HFD conditions, the Tm6sf2^167K/K mice had notably higher percentages of monounsaturated TG in monounsaturated lipids and polyunsaturated TG in polyunsaturated lipids (Fig. 2H), but lower percentage of monounsaturated or polyunsaturated PC in total monounsaturated or polyunsaturated lipids compared with the Tm6sf2^167E/E mice (Fig. 2I). These findings collectively suggest that the TM6SF2 E167K variant might influence the transfer of hepatic unsaturated fatty acids (particularly PUFAs) between TG and PC, resulting in an increase in polyunsaturated TG content and a decrease in polyunsaturated PC content in the liver of HFD-induced mice.

To further explore the molecular mechanism by which the TM6SF2 E167K variant regulated the levels of polyunsaturated TG and polyunsaturated PC, we used the string protein interaction network to screen for the functionally related proteins with TM6SF2 (Supplementary Fig. 4A). It has previously been shown that the deletion of PNPLA3 resulted in a decrease in hepatic polyunsaturated PC content and an increase in polyunsaturated TG content in mice [24,26]. This phenomenon is highly consistent with the alterations in hepatic polyunsaturated PC and TG levels induced by the TM6SF2 E167K variant that we identified in mice. We subsequently analyzed the relative expression levels of TM6SF2 and PNPLA3 genes in the livers of 94 NAFLD patients using the GEO database dataset (GSE174478), and found a strong positive correlation between the expression levels of TM6SF2 and PNPLA3 (Supplementary Fig. 4B). Next, we extracted primary hepatocytes from the Tm6sf2^167E/E and Tm6sf2^167K/K mice and treated them with FFAs. Immunofluorescence (IF) analysis showed a significant colocalization of TM6SF2 and PNPLA3 proteins, and the TM6SF2 E167K variant increased the co-localization of these proteins (Fig. 3A). Co-immunoprecipitation (Co-IP) results showed an interaction between TM6SF2 and PNPLA3 proteins under conditions with and without FFAs, and the TM6SF2 E167K variant significantly enhanced the interaction between TM6SF2 and PNPLA3 (Fig. 3B and Supplementary Fig. 5A). We overexpressed Myc-TM6SF2^167E or Myc-TM6SF2^167K plasmids in 293T cells. IF results also showed a co-localization between TM6SF2 and PNPLA3 proteins and the TM6SF2 E167K variant increased the co-localization (Supplementary Fig. 6). Co-IP analysis also indicated an interaction between TM6SF2 and PNPLA3, which was strengthened by the TM6SF2 E167K variant (Fig. 3C and Supplementary Fig. 5B). TM6SF2 is primarily distributed in the ER and PNPLA3 is distributed in the lipid droplets (LDs) [11,30]. We next explored whether the TM6SF2 E167K variant could affect the distribution of PNPLA3 in the LDs. Firstly, we extracted protein of the ER and LDs from primary hepatocytes of the Tm6sf2^167E/E and Tm6sf2^167K/K mice. Western blotting (WB) analysis indicated that the TM6SF2 E167K variant significantly enhanced PNPLA3 enrichment in the ER (Fig. 3D and Supplementary Fig. 5C) and reduced its distribution in the LDs in primary hepatocytes with or without FFA treatment (Fig. 3E and Supplementary Fig. 5D). In cells overexpressing Myc-TM6SF2^167E or Myc-TM6SF2^167K plasmids, the TM6SF2 E167K variant also enhanced PNPLA3 enrichment in the ER (Fig. 3F and Supplementary Fig. 5E) and vreduced its distribution in the LDs in cells with or without FFA treatment (Fig. 3G and Supplementary Fig. 5F). We further used the ER marker protein ERP72 and the LDs marker protein PLIN2 to fluorescently label the ER and LDs in primary hepatocytes from the Tm6sf2^167E/E and Tm6sf2^167K/K mice, respectively. IF results showed that the TM6SF2 E167K variant enhanced the colocalization of PNPLA3 with the ER, while conversely reducing its colocalization with the LDs in hepatocytes treated with FFAs (Fig. 3H and I). These results suggest that the TM6SF2 E167K variant could enhance the interaction between TM6SF2 and PNPLA3 proteins in the ER to cause the depletion of PNPLA3 in the LDs, which might influence PNPLA3’s biological functions within the LDs.

We also examined whether the PNPLA3 I148M variant affects the interaction between TM6SF2 and PNPLA3 proteins. Co-IP results showed that the PNPLA3 I148M variant did not cause significant changes in the interaction between TM6SF2 and PNPLA3 proteins (Supplementary Fig. 7). These results suggest that the PNPLA3 I148M variant could not affect the interaction between TM6SF2 and PNPLA3 proteins.

Previous studies have suggested that PNPLA3 is an enzyme involved in regulating lipids metabolism and could catalyze the transfer of PUFAs from TG to PC [24]. Based on our findings that the TM6SF2 E167K variant induces changes in the levels of PUFAs between TG and PC in the liver, it is reasonable to speculate that the TM6SF2 E167K variant might affect the transfer of PUFAs from TG to PC mediated by PNPLA3. We utilized alkynyl-labeled linoleic acid (C18:2) as a substrate and examined the changes in linoleic acid incorporated into TG or PC. The labeled lipids were separated on a thin-layer chromatography (TLC) plate and analyzed by fluorescence imaging to realize rapid analysis of newly synthesized lipids (Fig. 4A). Primary hepatocytes were extracted from the Tm6sf2^167E/E and Tm6sf2^167K/K mice and treated with FFAs and alkynyl-labeled linoleic acid. The results of linoleic acid incorporation assays showed that the newly synthesized PC/TG ratio in primary hepatocytes from the Tm6sf2^167K/K mice was significantly decreased compared with that in the Tm6sf2^167E/E mice, and the difference in the newly synthesized PC/TG ratio in primary hepatocytes between the two groups was significantly abolished after knockdown of PNPLA3 (Fig. 4B–D and Supplementary Fig. 8A). We also performed a confirmatory validation in 293T cells transfected with TM6SF2^167E or TM6SF2^167K plasmids. Upon overexpression of TM6SF2^167E or TM6SF2^167K plasmids, PNPLA3 knockdown was performed to validate the reversibility (Fig. 4E and Supplementary Fig. 8B). We observed a significant decrease in the PUFA-containing PC/TG ratio of newly synthesized lipids in cells with the TM6SF2 E167K variant compared with the control group (Fig. 4F and G), and the difference disappeared upon knockdown of PNPLA3 (Fig. 4H and I). These results indicate that the TM6SF2 E167K variant affects the function of PNPLA3, resulting in an elevation of polyunsaturated TG level and a reduction in polyunsaturated PC level.

PUFAs are particularly susceptible to the attack by oxygen free radicals, which induce lipids peroxidation stress and cellular damage [31]. ROS could induce lipid peroxidation, and MDA was a product of lipid peroxidation [32,33]. We treated primary hepatocytes from the Tm6sf2^167E/E and Tm6sf2^167K/K mice with FFAs and assessed the impact of the TM6SF2 E167K variant on the levels of intracellular ROS and MDA. Specifically, under FFA treatment conditions, both the ROS and MDA levels in primary hepatocytes from the Tm6sf2^167K/K mice were significantly elevated compared with those in the Tm6sf2^167E/E mice (Fig. 5A and B). In cells overexpressing Myc-TM6SF2^167E or Myc-TM6SF2^167K plasmids, we found that the TM6SF2 E167K variant could significantly increase the levels of ROS and MDA under FFA treatment (Fig. 5C and D). These findings indicate that the TM6SF2 E167K variant might exacerbate intracellular lipids peroxidation level of FFA-treated hepatocytes. We also investigated cell-membrane fluidity using a lipophilic pyrene probe called pyrene decanoic acid (PDA) (Fig. 5E). Our findings showed that FFA treatment significantly reduced cell-membrane fluidity in both mouse primary hepatocytes and 293T cells. The TM6SF2 E167K variant showed a noticeable enhancement of FFA-induced reduction in cell-membrane fluidity (Fig. 5F and G). Furthermore, the electron microscopy results of liver tissue in the HFDtreated mice showed that the Tm6sf2^167K/K mice exhibited more severe disruption and fragmentation of the ER, suggesting that the E167K variant could increase the sensitivity of the ER to HFD-induced lipid stress in hepatocytes (Fig. 5H). These data suggest that the TM6SF2 E167K variant significantly enhances lipid-induced MDA and ROS levels and exacerbates damage to the cell membrane of hepatocytes.

Using volcano plot analysis, we identified differentially expressed lipid molecules in the liver tissue between the Tm6sf2^167E/E and Tm6sf2^167K/K mice (Fig. 6A and B). Figure 6C shows the levels of differentially expressed phospholipid molecules downregulated by the TM6SF2 E167K variant or HFD in the liver tissue of mice. Here, we focused on PC containing C18:3, namely PC (16:0/18:3) +H and PC (18:3/20:4) +H for further analysis (Fig. 6D and E), which are two molecules with clear chemical structure and high abundance. The level of PC (16:0/18:3) was lower in the Tm6sf2^167K/K mice fed with either CD or HFD (Fig. 6D). Therefore, we speculated that the decreased level of PC containing C18:3 might represent a phenotype of MASLD with the TM6SF2 E167K variant and its deficiency might exacerbate hepatic steatosis.

To investigate whether the supplementation of PC containing C18:3 through diet could affect HFD-induced hepatic steatosis, we introduced dietary intervention with PC containing C18:3 in the HFD-induced Tm6sf2^167E/E or Tm6sf2^167K/K mice (Fig. 7A). Our results showed that the intervention with PC containing C18:3 abolished the increase in body weight, liver weight, hepatic index, hepatic MDA level, hepatic steatosis and hepatic TG content, and eliminated the reduction in membrane fluidity of hepatocytes caused by TM6SF2 E167K variant in HFD-fed mice (Fig. 7B–I). We also found that intervention with PC containing C18:3 abolished the increase in hepatic injury indicators including MASLD activity score, plasma ALT level, and plasma AST level induced by TM6SF2 E167K variant in HFD-fed mice (Fig. 7J–L). These findings suggest that PC containing C18:3 could attenuate HFD-induced hepatic steatosis and injury, and this attenuation effect is significantly dependent on the TM6SF2 E167K variant.

In this study, we identified significantly reduced content of PC (16:0/18:3) in the liver tissue of the HFD-induced Tm6sf2^167K/K mice. To investigate the clinical value of PC (16:0/18:3), we collected plasma samples from 21 MASLD patients and 11 HC. Table S4 presents the specific genotype and relevant clinical information.

We conducted targeted lipidomics analysis on the plasma samples obtained from HC and patients with MASLD. A total of 19 lipid subclasses and 879 lipid molecules were analyzed (Supplementary Fig. 9). The MASLD patients with TM6SF2^167E/K+K/K exhibited a significant decrease in plasma PC level compared with the MASLD patients with TM6SF2^167E/E (Fig. 8A). The MASLD patients with TM6SF2^167E/K+K/K showed a significant decrease in the level of polyunsaturated PC compared with the MASLD patients with TM6SF2^167E/E (Fig. 8B). The plasma TG level and the level of polyunsaturated TG showed no significant differences between the TM6SF2^167E/E and TM6SF2^167E/K+K/K groups (Supplementary Fig. 10). Through volcano plot analysis, we identified a set of lipid molecules with significantly decreased expression in the TM6SF2^167E/K+K/K group (Fig. 8C and D). Figure 8E showed the levels of various PC containing C18:3, and the level of PC (16:0/18:3), which was a differentially low-expressed PC identified in Tm6sf2^167K/K mice, was also lower in the patients with TM6SF2^167E/K+K/K than in those with TM6SF2^167E/E. In the MASLD patients correlation analysis revealed a significant negative correlation between the plasma PC (16:0/18:3) level and the liver fat content (r=−0.6143, P<0.05) (Fig. 8F). In the MASLD patients with TM6SF2^167E/K+K/K, the r value between the plasma PC (16:0/18:3) level and the liver fat content was up to −0.7095 (P<0.05) (Fig. 8G). However, in the MASLD patients with TM6SF2^167E/E, there was no significant correlation between the plasma PC (16:0/18:3) level and the liver fat content (r=−0.2193, P>0.05) (Fig. 8H). The significance of the plasma PC (16:0/18:3) level in predicting the TM6SF2 E167K variant was evaluated using ROC curve analysis. The area under the ROC curve (AUC) was 0.7636 to distinguish MASLD patients with TM6SF2^167E/K+K/K from those with TM6SF2^167E/E (Fig. 8I). The significance of plasma PC (16:0/18:3) level in predicting MASLD was also assessed using ROC curve analysis, and the AUC of plasma PC (16:0/18:3) level in predicting MASLD from HC was 0.7446 (P<0.05) (Fig. 8J).