Modeling Transmembrane 6 Superfamily Member 2 p.Glu167Lys-associated Steatotic Liver Disease Using Androgen-treated Human Induced Pluripotent Stem Cell-derived Hepatocyte-like Cells

Abstract

Background and Aims

Transmembrane 6 superfamily member 2 (TM6SF2) p.Glu167Lys (E167K) is associated with hepatic steatosis and male-predominant type 2 diabetes (T2D) presumably due to defective secretion of very low density lipoprotein (VLDL) particles. The causal relationship between secreted VLDL density and insulin sensitivity in the context of the E167K variant remains unclear.

Methods

E167K-carrying induced pluripotent stem cells were generated using a genome-editing method. Guided by targeted phospholipidomic analysis, we established an improved protocol to derive hepatocyte-like cells (HLCs) from induced pluripotent stem cells with significantly improved VLDL secretory profiles after linoleic acid, arachidonic acid and Liver X receptor agonist exposure. Furthermore, the E167K-defined steatosis and metabolic phenotype of HLCs treated with or without dihydrotestosterone were analyzed. In public database, we stratified male subjects by testosterone level and analyzed the impact of TM6SF2 on the incidence of hepatic steatosis and T2D.

Results

E167K hepatocytes showed defective VLDL secretion and an exacerbated lipid accumulation phenotype, accompanied by a decrease in arachidonic acid-containing phosphatidylcholine. Furthermore, dihydrotestosterone supplementation worsened the steatotic phenotype of E167K HLCs without altering insulin sensitivity. In the clinical data, E167K-defined hepatic steatosis was testosterone-dosage dependent unlike the other genetic variant, while T2D was not.

Conclusion

We report an improved human hepatocyte differentiation model to investigate metabolic dysfunction with functional VLDL secretory profiles influenced by TM6SF2 gene variant. Our combinatorial analysis of in vitro and human data indicates that male testosterone modifies the effect of E167K on hepatic steatosis but not insulin sensitivity. These data support the hypothesis that male-predominant T2D in E167K carriers is not causative but rather a consequence of persistent steatosis.

Graphical abstract

Introduction

Metabolic associated steatotic liver disease (MASLD) is a global epidemic with an increasing prevalence.¹ Genetic predisposition plays a significant role in the development and progression of MASLD including rs58542926 (Transmembrane 6 superfamily member 2 [TM6SF2] p.E167K), which is a major risk factor for steatotic liver disease.² In E167K carriers, increased hepatic lipid accumulation is accompanied by reduced serum lipid levels³ and impaired secretion of large, triglyceride (TG)-rich very low density lipoprotein (VLDL) particles.⁴ Boren et al. found that in E167K carriers, the production of large TG-rich VLDL decreased by 45%, whereas the production of smaller, less TG-rich VLDL increased by 15%, as measured by a deuterium-based VLDL kinetics assay.⁴ Similarly, Tm6sf2 knockout mice exhibit hepatic steatosis with impaired large VLDL secretion, mirroring the phenotype seen in humans carrying the E167K variant.^5,6 These reports indicate a role of the E167K variant in defective VLDL particle assembly or secretion, contributing to hepatic steatosis.

The molecular mechanisms by which the E167K variant leads to hepatic steatosis with reduced VLDL secretion remain unclear. In experiments using a human hepatoma cell line overexpressing TM6SF2, the E167K variant led to reduced TM6SF2 protein expression due to increased protein instability.⁷ Among the lysophosphatidylcholine acyltransferase (LPCAT) family proteins, LPCAT3 has the highest activity for incorporating n6–polyunsaturated fatty acid (PUFA; mainly arachidonic acid [AA]) into phosphatidylcholine (PC) and phosphatidylethanolamine (PE).⁸ LPCAT3 deficiency in the liver causes hepatic steatosis with impaired large VLDL secretion.^9,10 Therefore, AAs in phospholipids are believed to be essential for large VLDL secretion in hepatocytes. In E167K carriers, there are reduced levels of PUFAs in PC of extracted liver tissue,¹¹ suggesting reduced PUFA, especially AA-containing PC, may be the cause of the impaired large VLDL secretion. However, there is currently limited data at the human hepatocyte level to directly support a decrease in AA-containing phospholipids in E167K-carrying hepatocytes.¹¹

Recently, gene-gender interaction in MASLD has drawn attention for its potential to alter the functional significance of risk alleles.¹² For example, male carriers of the E167K variant have been reported to preferentially develop type 2 diabetes (T2D).¹³ Androgens may influence membrane homeostasis by reducing cholesterol and saturated fatty acids in phospholipids within biological membranes,¹⁴ as well as through the regulation of phospholipase A2 activity.¹⁵ Since TM6SF2 may induce changes in phospholipid composition,¹¹ there may be a possible link between androgen-induced alterations in membrane lipid composition and hepatic steatosis. However, the effects of androgens on hepatocytes carrying the E167K variant have not yet been investigated.

To date, hepatocellular carcinoma cell lines have been often used in previous studies to analyze the molecular functions of TM6SF2.^7,16,17 Although hepatoma cell lines are expandable and easy to genetically manipulate, the secreted lipoproteins are largely LDL-like.¹⁸ This may be a drawback in studying lipid accumulation related to TM6SF2 variants. While induced pluripotent stem cells (iPSCs)-derived hepatocyte-like cells (HLCs) offer an alternative model that secretes lipoproteins,¹⁹ it remains unclear whether the secreted VLDL size differs based on genetic risk variants. The synthesis of AA, an important membrane component for VLDL secretion,^9,10 is limited in human cells.²⁰ However, commonly used hepatocyte culture media are fatty acid-free.²¹ Therefore, revising conventional hepatocyte differentiation methods is crucial for studying VLDL secretion profiles by altering essential fatty acid composition.

Here, we took advantage of a phospholipidomic profiling approach to find key differences between primary human liver and conventional HLC system, and focused on the role of membrane-associated PUFA in the VLDL secretory process.^10,22,23 By supplementing PUFA with a liver X receptor (LXR) agonist to the HLC differentiation medium, we aimed to mimic the membrane lipid composition of primary hepatocytes to enhance VLDL secretion activity. Under these enhanced conditions, we assessed the effect of E167K by comparing the genome-edited cells with their isogenic counterpart. Further, we examined lipid accumulation and insulin sensitivity in dihydrotestosterone (DHT)-treated, E167K-carrying HLCs and analyzed clinical data to assess the impact of androgens on T2D and hepatic steatosis risk in E167K-carrying males.

Methods

Human iPSC Lines

The human iPSC line 1383D6 was obtained from Center for iPS Cell Research and Application, Kyoto University and used for this study with the approval by the Ethics Committee at Faculty of Medicine of Institute of Science Tokyo. The human iPSC lines YD8-1, YD5-1, and YD3-1 were established from the healthy donor as described previously.²⁴ Detailed methods can be found in Supplementary Information.

Results

Isogenic Derivation of TM6SF2 E167K-edited iPSC-derived Hepatocyte-like Cells

To examine the effect of TM6SF2 genotype on metabolic phenotypes in hepatocytes, we first generated the TM6SF2 E167K-mutated lines in a wild-type (EE) human iPSC by introducing a mutation of c.499G>A p.E167K in the TM6SF2 locus. Here we used a female donor-derived iPSC without the presence of other variants predictive of MASLD (PNPLA3 rs78409 and GCKR rs1260326) (Figure A1A). To study the zygosity-dependency of the variant, we applied the microhomology-assisted excision method (Figure 1A, Figure A1B and C).²⁵ We successfully derived the TM6SF2^E167K-homozygous (KK; c.499A/c.499A) and -heterozygous (EK; c.499G/c.499A) subclones as well as EE (c.499G/c.499G) (Figure 1B and Supplementary Methods). After battery of quality control testing in TM6SF2 EE, EK, and KK iPSC lines (Figure A1D–F), the TM6SF2-edited iPSC lines were then differentiated into HLCs using the previously published protocol²⁶ (Figure A1G). Time-course analysis using the reference iPSC line 1383D6 with TM6SF2 EE genotype revealed that TM6SF2 expression is induced at the early hepatic progenitor stage and maintained during day 12–18 of the hepatocyte maturation (Figure A1H). Hepatic gene expression including ALB, CYP3A7, CYP7A1, APOB, MTTP, and ASGR1 increased along with the induction of HLC differentiation. HLCs from either TM6SF2 genotype displayed a monolayer of polygonal-shaped epithelial cells (Figure 1C), which were positive for albumin, CK18, and ZO1 (Figure 1D). Quantitative PCR analysis revealed that these HLCs expressed comparable levels of hepatocyte marker genes including ALB, ASGR1, HNF4A, CEBPA, and NR1H3 (Figure 1E), as well as genes critical for the VLDL secretion such as APOB, MTTP, TM6SF2, and LPCAT3 (Figure 1F). Immunoblotting of HNF4α, FOXA2, and ALB proteins showed consistent expression profiles among the three TM6SF2-edited HLCs (Figure 1G). These HLCs secreted a comparable amount of albumin, which corresponds to approximately 6.0% of that of Primary Human Hepatocyte (PHH; Figure 1H). Thus, HLC differentiation efficiency was comparable across the three different genotype groups.

Figure 1.

Generation of TM6SF2 E167K-edited iPSC-derived hepatocyte-like cells. (A) Schematic illustration of the gene-targeting strategy for the TM6SF2-rs58542926 variant using the MhAX method. Homozygous (KK), heterozygous (EK), and wild-type (EE) iPSC lines were generated from a wild-type human iPSC line. (B) Sanger sequencing was performed to examine whether the human iPSC clones were correctly targeted. (C) Phase contrast images of genome-edited HLCs are shown. (D) Genome-edited HLCs were subjected to immunostaining with anti-ALB, anti-ZO1 and anti-CK18 antibodies. Nuclei were counterstained with Hoechst 33,342. (E and F) Gene expression levels of hepatic marker genes (ALB, ASGR1, HNF4A, CEBPA, and NR1H3) and VLDL assembly-related genes (APOB, MTTP, TM6SF2, and LPCAT3) in genome edited HLCs were examined by qRT-PCR (∗P < .05; Tukey’s post hoc tests). (G) Protein levels of HNF4α, FOXA2, Albumin, and α-tubulin in genome-edited HLCs were examined by western blotting. (H) Amounts of ALB secretion in genome-edited HLCs and PHHs were examined by ELISA (Tukey’s post hoc tests). All data represent mean ± SD. n.s., not significant.

Arachidonic Acid, Linoleic Acid, and LXR Agonist Exposure Facilitates VLDL Secretion in HLCs

Within hepatocytes, VLDL particles are assembled from TG with the aid of phospholipids, predominantly PC,²⁷ which form a hydrophilic coating monolayer. We performed LC-MS/MS-based, targeted phospholipid profiling of cell lysates to compare the fatty acid composition of phospholipids between HLCs derived from the reference iPSC line 1383D6, PHH and human liver tissue. Compared to PHH and liver tissue, HLCs cultured in the control condition (control HLCs) contained a relatively higher proportion of saturated and monounsaturated PC and PE while fewer polyunsaturated PCs and PEs, particularly those composed of fatty acids with four double bonds (Figure A2A). Control HLCs contained lower proportion of linoleic acid (18:2) (LA) and (20:4) (AA) in PC and PE than that in PHH (approximately 19.3% LA and 14.4% AA in PC and 43.9% LA and 33.2% AA in PE relative to those in PHH) (Figure A2B).

The incorporation of PUFA into PC directs the loading of TG termed lipidation upon the VLDL assembly and secretion in hepatocytes.^10,23 Several reports have shown that human iPSC-derived HLCs can synthesize and release ApoB100 and ApoE proteins in vitro,^28,29 but their ability to fully lipidate ApoB100 for the assembly of TG-rich VLDL has not been well characterized. In our HLC differentiation protocol, neither LA nor AA was included in the hepatocyte maturation medium (Figure A2C). We therefore supplemented LA and AA in the HLC culture (Figure 2A). Density gradient ultracentrifugation of the culture supernatants revealed that conditions supplemented with AA and LA slightly increased the levels of ApoB100 in the first fraction (fraction 1) (Figure 2A), which is equivalent to the large VLDL fraction with the highest TG over ApoB100 ratio in normal human plasma (Figure A2D). To enhance the incorporation of AA and LA into PC and facilitate large VLDL secretion, we focused on the LXR signaling. Synthetic LXR agonists have been reported to enhance hepatic VLDL production.³⁰ The LXR target genes LPCAT3¹⁰ and PLTP²² promote membrane phospholipid unsaturation and phospholipid transfer to VLDL, respectively. LXR also increases the synthesis of TG in liver, which is loaded into VLDL. The combination of a LXR agonist T0901317 (T) with AALA (termed AALAT) further increased the secretion of ApoB100 in fraction 1 (Figure 2A), leading to the similar profile as lipidated ApoB100 in human plasma (Figure 2B). The profile observed in AALAT HLCs was distinct from that of a hepatoma cell line HepG2, which secretes LDL-like lipoproteins owing to the limited ApoB100 lipidation.¹⁸ The total production levels of ApoB100 in the culture supernatants were similar between control and AALAT HLCs, both of which were higher than those observed in PHH (Figure 2C). In addition, HLCs treated with AALAT maintained their morphology and expression of albumin, CK18, and ZO1 (Figure A2E).

Figure 2.

Arachidonic acid, linoleic acid, and LXR agonist exposure facilitate VLDL secretion in HLC. (A) Schematic overview of the AALAT protocol and lipoprotein separation using iodixanol gradient ultracentrifugation. The bar graph shows the effect of AALA and T0901317 treatments on the amount of VLDL in Fraction 1 (∗P < .05; Tukey’s post hoc tests). (B) Upper: Lipoproteins were isolated from the culture supernatants of control HLCs, AALAT-HLCs, and HepG2 cells, and ApoB100 levels were measured by ELISA (n = 3 per a group; ∗P < .05; Student’s t-test). Lower: The ratio of triglyceride to ApoB in each fraction from human plasma is shown (n = 6 per group). (C) ApoB100 secretion of control HLCs, AALAT HLCs and PHH was analyzed (∗∗∗P < .001, Tukey’s post hoc tests). (D) Hierarchical clustering was performed for the ratio of fatty acid content in PC of control HLC, AALAT HLC and human liver tissue. (E) The TG content of control HLCs and AALAT HLCs was analyzed (∗∗∗P < .001, Student’s t-test). (F) Gene expression levels related to lipoprotein assembly and de novo lipogenesis were measured by qRT-PCR in control-, AALA-, and AALAT-HLCs. All data represent mean ± SD.

Hierarchical clustering based on the fatty acid composition in PC revealed that AALAT HLCs displayed a signature close to human liver and PHH, which is distinct from control HLCs (Figure 2D). Upon AALAT treatment, the levels of LA (18:2) incorporated into PC and PE in HLCs increased to approximately 59.2% and 86.1% of those in PHHs, respectively. The levels of AA (20:4) increased to approximately 44.5% and 61.2%, respectively (Figure A2A and B). Conversely, the amounts of monounsaturated PC and PE were decreased (Figure A2A). Intracellular accumulation of TG was increased in the AALAT condition (Figure 2E).

We next assessed gene expression under AALA and AALAT conditions by focusing on the LXR target genes. Quantitative PCR analysis revealed that AALAT, but not AALA, condition markedly upregulated the genes involved in VLDL assembly (MTTP, APOC1, PLTP, and LPCAT3) and fatty acid synthesis (SREBP1, FASN, SCD1, and ACC1) (Figure 2F),^22,23 indicating that the addition of LXR agonist augments these pathways in HLCs. Together, the combination of AA, LA, and an LXR agonist facilitates VLDL secretion in HLCs.

TM6SF2 E167K-edited HLC Exacerbates Steatosis Phenotype with Impaired VLDL Secretion

We next examined the impact of TM6SF2 E167K carriage on lipid metabolism by comparing isogenic TM6SF2 E167K-edited iPSC-derived HLCs. BODIPY 493/503 staining revealed significant lipid droplet accumulation in KK HLCs under AALAT but not control conditions (Figure 3A). Quantification of BODIPY intensity showed approximately a 1.65- and 1.34-fold increase in the TM6SF2 KK cells compared to EE and EK, respectively (Figure 3B). Intracellular TG levels were higher in KK cells under AALAT condition (Figure 3C). To assess VLDL secretory capacity across different TM6SF2 genotypes, we performed a time-course analysis of ApoB100 secretion. ApoB100 secretion in KK HLCs was reduced by 29.4% and 17.9% at 6 and 24 hours, respectively, compared to EE HLCs (Figure 3D). Density gradient profiling of ApoB100 in the culture supernatants revealed that, in contrast to EE HLCs, KK HLCs displayed an approximately 38.2% decrease in ApoB100 levels of the fraction 1 with a concomitant increase in the higher density fractions (2–3) (Figure 3E). Immunoblotting showed no significant difference in intracellular ApoB100 protein levels between EE and KK HLCs (Figure A3A), indicating that TM6SF2 E167K impairs the secretion of lipidated ApoB100, but not the de novo protein synthesis of ApoB100.

Figure 3.

TM6SF2 E167K-edited HLC exacerbates steatosis phenotype with impaired VLDL secretion. (A) Representative images of genome-edited HLCs ± AALAT treatment, stained for lipid droplets and nuclei. Lipid droplets were stained with BODIPY 493/503, and nuclei were counterstained with Hoechst 33,342. (B) The intensity ratio of BODIPY to nuclei in genome-edited HLCs was quantified (∗∗∗P < .001, ∗P < .05, Tukey’s post hoc tests). (C) The TG content of genome-edited HLCs was analyzed (∗∗∗P < .001, Tukey’s post hoc-tests). (D) ApoB100 secreted by genome-edited HLCs into fresh culture medium was measured at three time points (n = 4 per a group, ∗P < .05, Student’s t-test). (E) Lipoproteins were fractionated from the culture supernatants of EE or KK HLCs, and ApoB100 in each fraction was analyzed by ELISA (n = 4 per a group, ∗∗P < .01, ∗P < .05, Student’s t-test). (F) The number of double bonds in PC of EE or KK HLCs was examined by using LC-MS/MS (∗∗P < .01, Student’s t-test). (G) Fatty acid content in PC of EE or KK HLCs was analyzed (∗∗P < .01, Student’s t-test). (H) Schematic overview of the TG composition analysis is shown. Lipids were extracted from the culture supernatant and cell lysate, and analyzed using LC-MS/MS. (I) Principal component analysis was performed on the relative abundance of TG species inside and outside of EE or KK HLCs. (J) Heat map shows the changes in the proportion of TG species in KK HLC (∗P < .05, Student’s t-test). All data represent mean ± SD.

To compare the phospholipid profiles, we performed the targeted phospholipidomic analysis of EE and KK HLCs cultured under AALAT condition. The compositions of distinct phospholipid classes were not altered between genotypes (Figure A3B). The ratios of lysophosphatidylcholine to PC and of lysophosphatidylethanolamine to PE were also consistent between EE and KK HLCs (Figure A3C), indicating no apparent changes in phospholipid substrate metabolism. However, PC with over four double bonds of fatty acyls was found to be decreased in KK HLCs with a 32.1% reduction in AA content, while a concomitant increase in saturated PC (Figure 3F,G). No differences in the number of double bonds or PUFA content were observed in PE (Figure A3D and E). In addition, such changes in AA-containing PC molecular species were not observed in undifferentiated iPSC lines (Figure A3F and G), suggesting that the effect of TM6SF2 E167K on PC composition is specific to HLCs.

Given the reduction in AA incorporation into PC and in VLDL secretion in KK HLCs, we next determined the composition of fatty acids in TG by the targeted lipidomics approach using both the cell lysate and the culture supernatant. 176 TG species were detected in intracellular fraction; of which, 145 were also identified as extracellular components (Figure 3H). Principal component analysis analysis showed a distinct distribution of TG species between EE and KK HLCs both in intracellular and extracellular fractions (Figure 3I). Closer investigation of fatty acids composed in TG species revealed that relative abundance of unsaturated fatty acids with 1-5 double bonds tends to increase in KK HLCs, which was more evident in intracellular TG than in extracellular TG (Figure 3J, Figure A3H). These data highlight the impact of the TM6SF2 E167K variant on the unsaturated fatty acid distribution in PC and TG, the VLDL secretion, and intracellular lipid droplets in human hepatocytes.

To assess LXR activity in EE and KK HLCs, we measured the mRNA levels of LXR target genes known to play key roles in fatty acid synthesis (SREBP1, FASN, ACC1, and SCD1), phospholipid remodeling (LPCAT3), lipoprotein clearance (APOC1), bile acid synthesis (CYP7A1, and CYP27A1), and LXR itself (NR1H3).^10,31,32 Quantitative PCR analysis revealed that the expression levels of these genes were comparable between EE and KK HLCs, suggesting that differences in LXR activity do not account for the lower VLDL secretion in KK HLCs (Figure 1E and F, Figure A3I).

Dihydrotestosterone Treatment Augments the Steatosis Phenotype in TM6SF2 E167K-edited HLC Dissociated from Insulin Sensitivity

A recent report using sex-stratified GWAS as well as knock-in mouse models has demonstrated that TM6SF2 E167K carriers exhibit impaired glucose tolerance and increased susceptibility to T2D in a male-specific manner.¹³ This led us to investigate whether sex steroid hormones such as testosterone mediate the sex-specific effects of the TM6SF2 E167K variant in hepatocyte metabolic dysfunction. We first assessed the effects of DHT in AALAT HLCs derived from two different male donors, with YD5-1 and YD8-1 harboring EE and KK genotypes, respectively (Figure 4A, Figure A4A). BODIPY staining revealed that KK donor-derived HLCs accumulated more lipid droplets than EE cells in the AALAT condition. Addition of DHT increased the lipid droplet content in KK donor-derived HLCs, but not in EE donor cells (Figure 4B and C). To evaluate the DHT effects in the isogenic background, we also used TM6SF2 EE and KK edited iPSC lines and found that DHT treatment led to greater lipid accumulation in TM6SF2 KK-, but not EE-, edited HLCs under the AALAT condition (Figure 4D and E).

Figure 4.

Dihydrotestosterone treatment augments the steatosis phenotype in TM6SF2 E167K-edited HLC dissociated from insulin sensitivity. (A) Schematic overview shows the protocol for HLCs differentiation with AALAT and 200nM DHT. (B, D, C, E) B and D show representative images of EE or KK HLCs ± DHT stained for lipid droplets and nuclei. Quantification for male (YD5-1, YD3-1) and female (YD8-1) donor HLCs is shown in C and E, respectively (∗P < .05, ∗∗∗P < .001, Tukey’s post hoc tests). (F) The enrichment plot shows the correlation with the male-biased gene set in the DHT-treated HLCs. (G) The arachidonic acid (20:4) content of PC in EE or KK HLCs ± DHT treatment was measured (∗∗P < .01, ∗P < .05, Student’s t-test). (H) The phosphorylation of Akt induced by insulin in DHT-treated EE or KK HLCs was measured by western blotting. Akt phosphorylation was measured after cells were starved overnight in insulin-free medium followed by 100 nM insulin for 20 minutes. Cells were also treated with OA and PA (200 μM each) for steatotic induction for 48 hours. The band intensity quantification data is shown in (I). (J) Genes related to sugar and lipid metabolism were analyzed to assess the effect of DHT on insulin sensitivity in KK HLCs. After overnight insulin starvation, the cells were treated with 100 nM insulin, and RNA was collected after 4 hours followed by bulk-RNA-seq. The TPM of the control condition was taken as 1.0, and the fold change was analyzed (Tukey’s post hoc tests). All data represent mean ± SD. OA, oleic acid.

Given that DHT affects lipid droplet accumulation in TM6SF2 KK HLCs irrespective of donor sex, we next asked whether DHT directly biases the transcriptomic state of HLCs toward that of the human male liver. We found in the TM6SF2 KK-edited HLCs that androgen-responsive genes, TSC22D1 and RHOU,^33,34 were increased in a concentration-dependent manner upon DHT treatment (Figure A4B). Based on the male-biased gene set found in the previous microarray analysis of human liver tissues,³⁵ RNA-seq revealed that TM6SF2 KK-edited HLCs treated with AALAT and DHT displayed more male-biased gene signatures than those treated with AALAT alone (Figure 4F).

We next asked if DHT affects phospholipid profiles in HLCs. Phospholipidomic analysis revealed that in TM6SF2 KK-edited HLCs, but not in EE-edited cells, the abundance of AA-incorporated PC was decreased by DHT (Figure 4G). These data indicate that testosterone signaling augments the defect in AA incorporation into PC observed in the TM6SF2 E167K-carrying HLCs.

We further assessed whether the combined treatment of AALAT and DHT affects insulin signaling in the HLCs. To this end, we established conditions to induce hepatic insulin resistance by treating the cells with palmitic acid (PA). Immunoblotting revealed that PA inhibited the insulin-dependent phosphorylation of Akt in HLCs (Figure A4C). Upon loading HLCs with free fatty acids (oleic acid and PA), insulin-induced phospho-Akt levels were comparable between TM6SF2 EE- and KK-edited HLCs treated with DHT (Figure 4H and I). The phospho-Akt levels were also unaltered in the absence of DHT (Figure A4D and E). RNA-seq analysis of KK-edited HLCs showed insulin-regulated genes related to glucose and lipid metabolism did not change with the presence or absence of DHT (Figure 4J), supporting the lack of any changes in insulin signaling despite enhanced lipid accumulation by DHT.

The Effect of TM6SF2-rs58542926 on the Incidence of Male Steatosis Associated with Testosterone Levels

Our observation of the combined effects of TM6SF2 E167K with DHT in the HLC model prompted us to evaluate the clinical impacts of testosterone levels in males. Accordingly, we studied the TM6SF2 missense variant rs58542926 coding E167K mutation through the genotype-phenotype association datasets from the All of Us cohort,³⁶ focusing on individuals aged 50 (around menopause) and above to exclude the potential influences of female hormones. Consistent with the previous study,¹³ the rs58542926 was found to be associated with T2D more in males (odds ratio = 1.24, 95% confidence interval [CI; 1.32, 1.17], P value = 2.34 × 10⁻¹³) than in females (odds ratio = 1.09, 95% CI [1.516, 1.04], P value = 5.50 × 10⁻⁴), in which significant threshold of 5 × 10⁻⁸ in a typical genome-wide association study meets only the association in males (Figure A5A). The association of rs58542926 with steatosis, blood TG, and total cholesterol replicated in both sexes of all age groups (Table and Figure A5B). We next assessed the effects of rs58542926 variant in different ranges of testosterone levels. The male population, whose plasma testosterone levels were measured, was divided into three groups according to their levels (ng/dl): less than 300, 300–500, and more than 500. These groups accounted for 30%, 48%, and 22% of the total, respectively. We found that the odds ratio for the diagnosis of hepatic steatosis with rs58542926 was accentuated in the group with more than 500 ng/dl (less than 300 group: odds ratio = 1.11, 95% CI [0.70, 1.76], 300–500 group: odds ratio = 1.38, 95% CI [0.96, 2.03], more than 500 group: odds ratio = 2.27, 95% CI [1.32, 3.93]) (Figure 5). On the other hand, there was no such trend for T2D due to rs58542926 (less than 300 group: odds ratio = 0.77, 95% CI [0.55, 1.09], 300–500 group: odds ratio = 0.94, 95% CI [0.71, 1.25], more than 500 group: odds ratio = 1.03, 95% CI [0.66, 1.60]). We also performed the same analysis for PNPLA3-rs738409. The risk of hepatic steatosis due to rs738409 was not additive in the group with more than 500 ng/dl (less than 300 group: odds ratio = 1.12, 95% CI [0.84, 1.49], 300–500 group: odds ratio = 1.40, 95% CI [1.09, 1.80], more than 500 group: odds ratio = 1.51, 95% CI [1.02, 2.25]). These findings indicate that testosterone is associated with hepatic steatosis, but not with T2D incidence in E167K-carrying males. This dissociated association may reflect the result that DHT-triggered lipid accumulation without insulin resistance in E167K HLCs.

Table.

Association Between rs58542926 and Metabolic Traits in Males and Females in All of Us

Quantitative traits
Sex	Male								Female
rs58542926	-			+			SNP +/−	P value	-			+			SNP +/−	P value
	Average	SD	N	Average	SD	N			Average	SD	N	Average	SD	N
TG (mg/dL)	133.67	72.50	44,551	125.71	67.15	4233	0.94	2.9E-13	121.06	63.48	73,077	114.71	58.80	6,778	0.95	3.5E-17
TC (mg/dL)	169.37	37.67	43,299	167.05	35.56	4143	0.99	6.7E-05	187.38	37.36	72,604	186.11	36.37	6,747	0.99	.0062
HDL (mg/dL)	47.93	14.11	42,741	48.58	14.62	4072	1.01	.0066	58.69	17.22	72,063	59.44	17.11	6,673	1.01	.0006
LDL (mg/dL)	95.54	31.38	42,417	94.42	30.15	4073	0.99	.0238	105.50	31.31	69,131	104.41	30.46	6,388	0.99	.0064
HbA1c (%)	6.30	1.53	31,799	6.24	1.49	2970	0.99	.0241	6.11	1.42	47,809	6.06	1.41	4,217	0.99	.0274
Glucose (mg/dL)	117.55	39.92	62,299	117.30	39.55	5732	1.00	.6472	109.48	35.79	105,288	108.81	34.73	9,645	0.99	.0705
BMI	28.92	6.41	116,309	28.71	6.04	10,365	0.99	.0007	30.32	8.07	178,851	29.77	7.78	16,096	0.98	1.2E-17
Age	57.99	17.06	142,962	59.04	17.13	10,617	1.02	1.4E-09	54.08	16.93	231,778	55.64	17.01	16,466	1.03	8.9E-30

Binary traits
Sex	Male			Female
	OR	N	P value	OR	N	P value
Steatosis	1.64	153,579	3.3E-32	1.64	248,244	2.4E-51
T2D	1.20	155,169	1.6E-11	1.14	249,565	2.4E-08
MI	1.05	153,579	0.246	1.23	248,244	2.5E-05
IHD	1.21	153,579	1E-06	1.14	248,244	.0014

The association of rs58542926 with the indicated metabolic traits was analyzed separately for genders. P values were analyzed between the presence and absence of the SNP.

BMI, body mass index; HbA1c, hemoglobin A1c; IHD, ischemic heart disease; LDL, low-density lipoproteins; MI, myocardial infarction; OR, odds ratio; SD, standard deviation; T2D, type 2 diabetes; TC, total cholesterol; TG, triglyceride.

Figure 5.

The testosterone level-stratified effect of TM6SF2-rs58542926 on the male steatosis and T2D. Forest plots show odds ratios for hepatic steatosis (top) and T2D (bottom) for TM6SF2 - rs58542926 (left) and PNPLA3 - rs738409 (right). The male population in which testosterone levels were measured was divided into three groups according to testosterone levels. Data represents OR ± 95% CI. The numbers in parentheses indicate the sample size.

Discussion

Here we established an HLC-based model to study the TM6SF2 E167K gene variant function on metabolic dysfunction under the PUFA-enriched culture condition. We demonstrated in the isogenically gene-edited HLCs that E167K inhibits the incorporation of AA into PC and VLDL secretion to extracellular milieu, concomitant with an increase in the intracellular TG. We also found an interaction of the TM6SF2 variant with testosterone both in vitro and in clinical cohort analyses. In the DHT-induced, masculinized HLCs carrying TM6SF2 E167K, the impairment of AA incorporation into PC as well as TG metabolism becomes more evident than in the reference allele-carrying cells, without disturbing insulin signaling. Consistent with the in vitro observations, analysis of the All of Us cohort revealed that the susceptibility to steatosis is positively influenced by plasma testosterone in TM6SF2 E167K male carriers. Collectively, our data provided insights into the interaction of the TM6SF2 E167K variant with testosterone in the male MASLD condition.

Recently, an iPSC-based model of E167K-associated hepatic steatosis was reported by Faccioli et al.³⁷ While they demonstrated that E167K-carrying HLCs lead to VLDL secretion deficiency, the VLDL measurement method did not characterize its density or TG content. This raised the question of what characteristics of VLDL metabolism were affected by the E167K mutation. By fractionating the HLC culture supernatant using density gradient ultracentrifugation, we showed that E167K specifically impairs the secretion of the lower density fraction of VLDL. The AALAT HLCs we utilized have potential applications in research on VLDL size and lipidation, extending beyond the scope of previous TM6SF2 studies using in vitro human hepatocyte models. Global lipidomics dataset by Faccioli et al showed that various lipid classes, including TG, diacylglycerol, phospholipids, and sphingolipids, increased in E167K HLC, leading to hepatic ER stress. In contrast, our targeted lipidomics observed a reduction in AA-containing PC in E167K HLCs, which may be associated with impaired secretion of the lower density VLDL fraction.^9,10 This mechanistic change could contribute to the sequence of critical events involving ER stress, mitochondrial dysfunction, apoptosis, and lipid metabolism ,²³ as seen in past TM6SF2 studies.^13,37,38

MASLD is more prevalent in females carrying the PNPLA3-rs738409 allele, while sex differences in MASLD and liver injury related to TM6SF2 did not note any significant findings.³⁹ However, another report indicates an increased risk of T2D in E167K-carrying males,¹³ which can subsequently trigger steatotic liver disease.⁴⁰ In vitro modeling tool is powerful in dissociating phenotypic causation and consequence between two interrelated manifestations: hepatic steatosis and insulin resistance. Indeed, our in vitro HLC modeling shows the association between testosterone and steatosis but not insulin responsiveness in E167K-carrying hepatocytes. Currently, findings around TM6SF2 E167K and DHT were evaluated exclusively using an HLC model. Reproducing these findings in PHHs is important but remains challenging in part due to the low frequency of TM6SF2 homozygotes (∼0.5%, gnomAD) as well as the difficulty in genetic manipulations. These technical hurdles currently limit PHH-based validation.

Therefore, we extended our in vitro observations in clinical datasets by stratifying the male population based on testosterone levels and investigated the impact of TM6SF2 in each group. Consistent with in vitro experimental modeling, analysis of All of Us cohort indicates that higher androgen levels exacerbate the steatosis phenotype in a dosage-dependent fashion but do not alter T2D susceptibility. These results warrant future exploration to support the hypothesis that androgen-driven chronic hepatic steatosis related to TM6SF2 leads to hepatic insulin resistance as a secondary consequence.

MASLD is a complex metabolic disease involving signaling interactions between hepatic parenchymal and nonparenchymal cells. In experiments using hepatic stellate cell lines, TM6SF2 has been reported to suppress differentiation into myofibroblasts, while E167K affects hepatic stellate cell activation by enhancing the response to TGFβ.⁴¹ The E167K mutation is likely to reduce the expression levels of TM6SF2 protein, potentially enhancing liver fibrosis in SNP carriers. Indeed, several clinical studies have reported an increased risk of liver fibrosis associated with E167K carriers.^39,42 Genome-edited iPSCs and associated differentiation protocols developed in this study will be applicable to multicellular liver organoids that include nonparenchymal populations^24,43, 44, 45 with zonation feature,⁴⁶ serving as a valuable investigative tool for interrogating dissociated impacts on steatosis, metabolic dysfunction and inflammation.

Conclusion

Our study suggests that TM6SF2 E167K may impair arachidonic acid incorporation into PC and VLDL secretion, and that this dysfunction appears to be influenced by testosterone. These findings provide potential mechanistic insights into the male-predominant susceptibility to MASLD and indicate the utility of genome-edited iPSC-derived hepatocyte models for exploring genetic variants and sex-related factors in liver disease.

Supplementary Materials

Figure A1.

Figure A2.

Figure A3.

Figure A4.

Figure A5.