Hepatocyte mARC1 promotes fatty liver disease

Abstract

Background & Aims

Non-alcoholic fatty liver disease (NAFLD) has a prevalence of ∼25% worldwide, with significant public health consequences yet few effective treatments. Human genetics can help elucidate novel biology and identify targets for new therapeutics. Genetic variants in mitochondrial amidoxime-reducing component 1 (MTARC1) have been associated with NAFLD and liver-related mortality; however, its pathophysiological role and the cell type(s) mediating these effects remain unclear. We aimed to investigate how MTARC1 exerts its effects on NAFLD by integrating human genetics with in vitro and in vivo studies of mARC1 knockdown.

Methods

Analyses including multi-trait colocalisation and Mendelian randomisation were used to assess the genetic associations of MTARC1. In addition, we established an in vitro long-term primary human hepatocyte model with metabolic readouts and used the Gubra Amylin NASH (GAN)-diet non-alcoholic steatohepatitis mouse model treated with hepatocyte-specific N-acetylgalactosamine (GalNAc)–siRNA to understand the in vivo impacts of MTARC1.

Results

We showed that genetic variants within the MTARC1 locus are associated with liver enzymes, liver fat, plasma lipids, and body composition, and these associations are attributable to the same causal variant (p.A165T, rs2642438 G>A), suggesting a shared mechanism. We demonstrated that increased MTARC1 mRNA had an adverse effect on these traits using Mendelian randomisation, implying therapeutic inhibition of mARC1 could be beneficial. In vitro mARC1 knockdown decreased lipid accumulation and increased triglyceride secretion, and in vivo GalNAc–siRNA-mediated knockdown of mARC1 lowered hepatic but increased plasma triglycerides. We found alterations in pathways regulating lipid metabolism and decreased secretion of 3-hydroxybutyrate upon mARC1 knockdown in vitro and in vivo.

Conclusions

Collectively, our findings from human genetics, and in vitro and in vivo hepatocyte-specific mARC1 knockdown support the potential efficacy of hepatocyte-specific targeting of mARC1 for treatment of NAFLD.

Impact and implications

We report that genetically predicted increases in MTARC1 mRNA associate with poor liver health. Furthermore, knockdown of mARC1 reduces hepatic steatosis in primary human hepatocytes and a murine NASH model. Together, these findings further underscore the therapeutic potential of targeting hepatocyte MTARC1 for NAFLD.

Graphical abstract

Highlights

• •Genetically predicted increases in MTARC1 mRNA associate with poor liver health.
• •mARC1 knockdown decreases lipid accumulation in primary human hepatocytes.
• •Hepatocyte-specific knockdown of mARC1 improves steatosis in a murine NASH model.

Introduction

Non-alcoholic fatty liver disease (NAFLD) encompasses a spectrum of histologically categorised pathologies including simple steatosis, non-alcoholic steatohepatitis (NASH), and cirrhosis.¹ Elucidating the impact of genetic variation on disease onset and progression has identified novel therapeutic targets for this major unmet medical need.² Genome-wide association studies (GWAS) identified and validated a common missense variant (p.A165T, rs2642438 G>A) in mitochondrial amidoxime-reducing component 1 (MTARC1) that protects against all-cause cirrhosis.[3], [4], [5] The minor A allele of rs2642438 is associated with decreased liver fat,^3,6 liver enzymes,^3,7 blood cholesterol,^3,8 and risk of NAFLD.^9,10 A low-frequency coding variant (p.M187K, rs17850677 T>A) and a rare stop codon variant (p.R200Ter, rs139321832 C>T) in MTARC1 are associated with lower blood cholesterol and protection from cirrhosis,³ suggesting mARC1 inhibition may have therapeutic potential for the treatment of liver disease. A slight increase in plasma triglycerides (TG) was observed in carriers of the minor A allele; however, cardiometabolic disease-based association studies of this variant in MTARC1^3,8,11 do not raise any concerns for mARC1 modulation. Effects of mARC1 beyond cardiometabolic disease have not been well captured, and trait-based phenome-wide association studies may inform mARC1 biology and its associated safety profile.

The molecular function of mARC1 within NAFLD pathophysiology is unclear. mARC1 is a molybdenum-containing enzyme anchored to the outer mitochondrial membrane,^12,13 where it catalyses the reduction of N-oxygenated substrates, including nitric oxide and xenobiotics,¹⁴ in combination with cytochrome b₅ type B and NADH cytochrome b₅ reductase.^15,16 Although hepatic mARC1 expression is regulated by nutritional stimuli such as glucose deprivation, fasting and high-fat diet,¹⁷ little is known regarding the metabolic consequences of altering mARC1 levels. Depletion of mARC2, a mARC1 paralogue, decreases formation of glycerolipids in adipocytes, and mARC2 knockout mice are protected from diet-induced obesity and associated metabolic disturbances.[18], [19], [20] As mARC1 is highly expressed in adipocytes, the sufficiency of targeting mARC1 in hepatocytes to confer the genetically predicted protective metabolic effects is unknown.

Here, we compile in silico, in vitro, and in vivo work to establish that hepatocyte MTARC1 is causally associated with NAFLD. We identify novel genetic associations with body composition and liver enzymes, demonstrate a shared genetic aetiology of NAFLD-related traits at the MTARC1 locus, and establish a causal association of genetically predicted MTARC1 expression with these traits. mARC1 knockdown in a long-term primary human hepatocyte (PHH) system showed decreased lipid accumulation. In a murine model of NASH, we found that hepatocyte-specific mARC1 depletion decreases steatosis and markers of fibrosis. Together, this work provides the first experimental evidence that inhibition of hepatocyte mARC1 is sufficient to reduce hepatic steatosis.

Materials and methods

Please see the Supplementary CTAT Table.

Genetic analyses

All analyses were performed in R v3.6.3 (R Foundation for Statistical Computing, Vienna, Austria) unless otherwise stated. Full details are provided in the Supplementary Materials and Methods and Tables S1–S3.

Multi-trait colocalisation

To assess whether the genetic associations reported with metabolic traits in the region of MTARC1 were attributable to the same underlying causal variant, statistical colocalisation using the Hypothesis Prioritisation in multi-trait Colocalisation (HyPrColoc) method was performed using priors, uniform priors, and sensitivity analyses using nonuniform priors (Supplementary Materials and Methods).

Mendelian randomisation

Two-sample Mendelian randomisation (MR) was performed to test the association of MTARC1 mRNA expression (Genotype-Tissue Expression tissues, blood, and immune and blood cells; Table S1) with metabolic outcomes (Table S2). The two-sample MR approach (compared with one-sample MR) allowed us to use summary association statistics from two distinct studies without the need for individual participant data. See the Supplementary Materials and Methods for details.

Rare variant association tests in the UK biobank

We tested the association of rare variants in the region of MTARC1 for association with the blood biomarkers (alanine aminotransferase [ALT], aspartate transaminase [AST], and alkaline phosphatase [ALP]) in 414,763 European ancestry participants in the UK Biobank with whole exome sequencing.²¹ Initial quality control (sex discordance, contamination, etc.) was performed by Regeneron Pharmaceuticals Inc²² and was complemented by our bioinformatic pipeline. See the Supplementary Materials and Methods for details.

PHH cell culture, siRNA transfection, and free fatty acid mixture

Cyroplatable PHH were thawed, plated, and transferred (24 h post thaw) to media containing five chemicals (5C) referred to as PHH maintenance media²³ (Tables S4 and S5). PHH were transfected on Day 8, treated with 0 or 800 μM free fatty acid (FFA) mix on Day 25, and harvested for RNA or fixed on Day 28. See the Supplementary Materials and Methods for details.

RNA extraction, cDNA synthesis, qPCR, and Western blotting

Details are provided in the Supplementary Materials and Methods.

Nile red staining

The Nile red ratio was calculated as the neutral lipid fluorescence (540–15 nm/600–20 nm) divided by the phospholipid fluorescence (540–15 nm/640–20 nm).²⁴ To normalise data across experiments, non-targeting siRNA (siNT) values were set to 0% for 0 μM FFA mix and 100% for 800 μM FFA mix treatment, using the following equation: % of lipid accumulation = ((Nile Red Ratio_{gene of interest} - Nile Red Ratio_{non-targeting 0 μM FFA})/(Nile Red Ratio_{non-targeting 800 μM FFA} - Nile Red Ratio_{non-targeting 0 μM FFA})) × 100. Details are provided in the Supplementary Materials and Methods.

Supernatant assays

Conditioned media were collected from PHH in a 96-well format after 72 h in culture (0 or 800 μM FFA mix). Apolipoprotein B (apo B) content was quantified using a commercial apo B assay (Cisbio, 64APBPEH). Media concentrations of TG and FFA were measured on the AU480 chemistry analyser (Beckman Coulter; High Wycombe, UK). Metabolomic profiles were assessed using the Metabolon HD4 platform as detailed in the Supplementary Materials and Methods.²⁵

In vivo mouse model

Two in vivo studies were carried out. Male C57BL/6JRj mice (n = 10–14 or n = 10–16), 5 weeks old, were fed the Gubra Amylin NASH (GAN) diet (40% fat, 22% fructose, and 2% cholesterol; D09100310, Research Diets, USA) for 44 or 36 weeks, respectively. Mice were dosed weekly s.c. for the last 8 weeks with 3 mg/kg N-acetylgalactosamine (GalNAc)-conjugated siRNA targeting murine Mtarc1 (GalNAc-siMtarc1). Data were compared with a NASH vehicle control group. All animal experiments were conducted in accordance with Novo Nordisk and Gubra bioethical guidelines, which are fully compliant to internationally accepted principles for the care and use of laboratory animals. The described experiments were covered by personal licenses for Jacob Jelsing (licenses #2013-15-2934-00784 and #2017-15-0201-01215) issued by the Danish Committee for Animal Research. See the Supplementary Materials and Methods for details and Table S6. Mouse plasma metabolites were analysed by MS-Omics using a Thermo Scientific Vanquish LC coupled to Thermo Q Exactive HF MS. An electrospray ionisation interface was used as the ionisation source in negative and positive ionisation mode. Ultra-performance liquid chromatography was performed using a slightly modified version of a previously described protocol.²⁶

Transcriptome (mRNA) library preparation and sequencing

Details of RNA isolation, library preparation, sequencing, read quality control, and alignment for the PHH and in vivo samples are provided in the Supplementary Materials and Methods.

Bioinformatic analyses of transcriptomic and metabolomic data

Details of the bioinformatic analyses are provided in the Supplementary Materials and Methods.

Statistical analysis of in vitro and in vivo experiments

GraphPad Prism 9.0.1 was used for statistical analysis of in vitro experiments. For lipid accumulation and apo B in PHH, two-way ANOVA for siRNA and FFA mix conditions with Tukey’s post hoc testing within FFA mix conditions was used. For PHH experiments performed in a single FFA mix condition, a paired two-sided t test was performed. Unpaired two-sided t tests were used to evaluate all in vivo results. Significance is shown as follows: ∗p ≤0.033, ∗∗p ≤0.002, and ∗∗∗p ≤0.001. No adjustment of p values to account for multiple comparisons has been made.

Results

Metabolic trait associations with MTARC1 are attributable to the same genetic variants

Despite data showing associations at the MTARC1 locus with ALT, AST, ALP, plasma LDL, total cholesterol (TC), apolipoprotein A (apo A), and apo B,³ it is unclear whether these associations are attributable to the same shared or distinct genetic causal variants and mechanisms. To address this, the HyPrColoc²⁷ method was used (Supplementary Materials and Methods). The genetic associations of liver enzymes (ALT, AST, ALP, and Sex hormone-binding globulin (SHBG)), circulating lipids (apo A, HDL cholesterol, apo B, LDL, and cholesterol), liver fat accumulation, and body composition traits (whole-body fat-free mass, whole-body water mass, trunk fat-free mass, and basal metabolic rate) identified at the MTARC1 locus were attributable to the same genetic variant, that is, colocalised with a posterior probability of 0.80 (Fig. 1A and Fig. S1). The missense variant (p.A165T, rs2642438 G>A) explained 100% of the posterior probability, suggesting it could be the causal variant (Fig. 1B). Thus, these data provide support for a shared genetic mechanism underlying the associations of these metabolic traits at the MTARC1 locus.

Fig. 1.

Statistical colocalisation of cardiometabolic traits at the MTARC1 locus.

(A) Regional association plots at the MTARC1 gene locus, showing the shared genetic associations with liver enzymes (ALT and AST), lipids (TC), and liver fat accumulation. Details about the statistical analysis and source of the data are given in the Materials and Methods section. Colour key indicates the correlation, r², with the respective lead variants in the GWAS. (B) Plot showing the colocalisation posterior probabilities explained by each of the genetic variants at the extended MTARC1 locus region in the colocalisation analysis. The table shows the parameters used for HyPrColoc analysis and results from HyPrColoc. ALT, alanine aminotransferase; AST, aspartate transaminase; GWAS, genome-wide association studies; HLX, H2.0 like homeobox; HyPrColoc, Hypothesis Prioritisation in multi-trait Colocalisation; MARK1, Microtubule Affinity Regulating Kinase 1; MTARC1, mitochondrial amidoxime-reducing component 1; TC, total cholesterol.

Rare and common exonic variants in MTARC1 associate with liver phenotypes

To systemically elucidate the impact of MTARC1 exonic variants on liver enzymes, single-variant association analysis of 270 exonic MTARC1 variants with ALT, AST, ALP, circulating lipids, and body composition traits were performed. We replicated the association of the G allele of the common missense variant (p.A165T, rs2642438 G>A) with increased ALT, ASP, and AST as well as with circulating lipids (Table S7). Further, we identified a novel association of a rare missense variant, rs12023067 A allele, with increased ALT (Table S7). Another rare missense variant, rs144056103, was suggestively associated with increased AST (p <5 × 10^-5) (Table S7). To improve statistical power to identify rare variant associations, we performed gene-based aggregate tests of rare MTARC1 variants with liver enzymes. No significant (p <1 × 10^-5) effects were obtained for ALP, ALT, AST, circulating lipids, or body composition traits (Table S8).

MTARC1 gene expression associates with metabolic traits

Genetic variants associated with MTARC1 gene expression (i.e. expression quantitative trait loci [eQTLs]) were used to test whether genetically determined MTARC1 gene expression is associated with metabolic conditions. Publicly available eQTL data for MTARC1 in 114 datasets were used, and up to three distinct eQTLs (distinct genetic variants, r² <0.1, associated with MTARC1 expression with p ≤5 × 10^-8) were identified, in 16 unique tissues or cells including muscle, left ventricle of the heart, stomach, transverse colon, and rectum (Supplementary Materials and Methods and Table S3A). Importantly, most of the identified instrumental variables (IVs) are not in linkage disequilibrium with the common missense variant (p.A165T, rs2642438 G>A) and show no association with MTARC2 gene expression (Table S3A and S3B). Analysis of these IVs showed that for every genetically predicted SD increase in MTARC1 gene expression levels in muscle, there is an associated increase in levels of ALT, AST, ALP, apo A, apo B, TC, and liver fat accumulation, but lower levels of basal metabolic rate and whole-body fat-free mass (Table 1 and Fig. S2). Similarly, genetically predicted increases in MTARC1 gene expression level in muscle were nominally associated with increased fibrosis and cirrhosis of liver (Table S9 and Fig. S2). The analyses of MTARC1 gene expression were performed in other tissues where eQTLs were available, and consistent results were obtained in the left ventricle and rectum with those obtained in muscle (Table S9). These results were robust to sensitivity analyses with a range of different MR approaches (Supplementary Materials and Methods). Although causal estimates from MR–Egger were less significant than those derived from the MR–inverse-variance weighting (IVW) approach, we obtained consistent effect directions of MTARC1 gene expression on liver phenotypes (Table S9). Taken together, our analyses suggest that genetically determined increases in MTARC1 gene expression are associated with adverse effects on liver enzymes, liver fat, and circulating lipids.

Table 1.

MR estimates for the effect of genetically determined gene expression level of MTARC1 on the metabolic traits.

Exposure	Outcome	MR causal estimate (IVW)		Heterogeneity
		Beta (SE)†	p value	p value‡
MTARC1 gene expression (muscle)∗	Alkaline phosphatase	0.0841 (0.0061)	2.96E-43	0.8102
	Alanine aminotransferase	0.0527 (0.0054)	2.08E-22	0.3087
	Aspartate aminotransferase	0.036 (0.0057)	2.03E-10	0.0011
	Direct bilirubin	0.0495 (0.0063)	6.37E-15	0.374
	Total bilirubin	0.0778 (0.006)	6.48E-38	0.0285
	SHBG	0.0306 (0.0057)	7.80E-08	0.2329
	Apolipoprotein A	0.0387 (0.0059)	6.45E-11	0.9309
	Apolipoprotein B	0.0445 (0.0059)	4.00E-14	0.1508
	Cholesterol	0.0469 (0.0057)	1.80E-16	0.9365
	LDL direct	0.045 (0.0057)	1.90E-15	0.9305
	Urate	-0.0409 (0.0095)	1.55E-05	0.3518
	Phosphate	0.0254 (0.0059)	1.41E-05	0.9591
	Liver fat accumulation	0.0648 (0.018)	4.58E-04	0.0655
	Whole-body fat-free mass	-0.021 (0.0039)	6.11E-08	0.0266
	Whole-body water mass	-0.021 (0.0039)	5.84E-08	0.0294
	Basal metabolic rate	-0.0193 (0.0041)	1.97E-06	0.0897
	Trunk fat-free mass	-0.0209 (0.0039)	8.93E-08	0.0574

Effect estimates and p values are provided for the IVW method. p-values are shown from the Cochran’s Q test for heterogeneity. Full details of the results from the different MR analyses, including details of data sources and number of cases, are reported in Tables S1, S2, S3A and B, and S9.

eQTL, expression quantitative trait locus; IVW, inverse-variance weighting; MTARC1, mitochondrial amidoxime-reducing component 1; MR, Mendelian randomisation.

^∗MTARC1 gene expression in muscle is instrumented by three single nucleotide polymorphisms (cis-eQTL).

^†Represents metabolic traits SD change per SD increase in MTARC1 gene expression in muscle.

^‡Heterogeneity p value was assessed using Cochran’s Q test.

mARC1 knockdown decreases lipid accumulation and apo B secretion in PHH

To assess the impact of MTARC1 loss of function in hepatocytes, an siRNA transfection protocol was established to knock down MTARC1 gene expression in PHH. PHH transfected with MTARC1 targeting siRNA (siMTARC1) maintained reduced MTARC1 mRNA levels compared with siNT for 17 days (Fig. 2A). Consistent with the 12-day predicted protein half-life of mARC1 in PHH,²⁸ decreases in mARC1 protein were observed 10 days following siMTARC1 treatment but became more pronounced on Day 17 (Fig. 2B and Fig. S3A). Protein and mRNA expression levels of mARC2 were unaffected by siMTARC1 (Table S3B and C). The effect of siMTARC1 and siRNA targeting Diacylglycerol O-acyltransferase 2 (DGAT2) (siDGAT2) on lipid accumulation was determined in the presence or absence of 800 μM FFA (Fig. 2C). The siRNAs effectively reduced expression (56–86%) of their target mRNA (DGAT2 and MTARC1) in 25 individual experiments using seven PHH donors (Table S10). Both siMTARC1 and siDGAT2 significantly reduced lipid accumulation in both conditions (Fig. 2D and E). Analysis of conditioned media from 16 experiments demonstrated that both siMTARC1 and siDGAT2 significantly decreased apo B secretion (Fig. 2F and Table S11). To assess the effect of mARC1 on lipid dynamics, siRNA-treated PHH were incubated with ¹³C-labelled glucose and fructose or ²H-labelled palmitate and oleate for the measurement of de novo lipogenesis (DNL) and fatty acid (FA) oxidation, respectively. mARC1 knockdown had no effect on DNL or FA oxidation (Fig. S4A and B). To determine whether mARC1 knockdown alters the import and export of lipids, FFA and TG concentrations were quantified in conditioned media from the above experiments. The majority of the 800 μM FFA was taken up by the PHH, and this was unaffected by siMTARC1 (Fig. S4C). In contrast, siMTARC1 significantly increased the concentration of TG in conditioned media (Fig. 2G).

Fig. 2.

mARC1 knockdown decreases lipid accumulation and apo B secretion in PHH.

Validation of (A) mRNA and (B) protein levels following MTARC1 siRNA knockdown in PHH over 17 days. (C) Workflow for culturing of PHH, siRNA transfection, and endpoint collections. (D) Lipid accumulation data following siRNA-mediated knockdown of MTARC1 and DGAT2 relative to NT siRNA, n = 25 independent experiments. (E) siRNA knockdown of MTARC1 and DGAT2 reduces lipid accumulation with and without FFA mix loading in PHH. Fluorescence images of nuclei (blue) and lipid droplets (hot) from siRNA knockdown PHH treated with FFA mix. Scale bar: 50 μm. (F) Reduced apo B expression observed following knockdown with MTARC1 siRNA relative to NT siRNA, n = 16 independent experiments. (G) Increased media TG levels observed following knockdown with MTARC1 siRNA relative to NT siRNA, n = 8 independent experiments. Data are presented as mean ± 95% CI, ∗p ≤0.033, ∗∗p ≤0.002, ∗∗∗p ≤0.001 (D and F, two-way ANOVA with Tukey’s post hoc testing within FFA mix conditions; G, paired t test). 5C, five chemicals; apo B, apolipoprotein B; FC, fold change; FFA, free fatty acid; KD, knockdown; MTARC1, mitochondrial amidoxime-reducing component 1; NT, non-targeting; PHH, primary human hepatocytes; RQ, relative quantification; siDGAT2, siRNA targeting DGAT2; siMTARC1, MTARC1 targeting siRNA; TBP, TATA-box binding protein; TG, triglycerides.

Hepatocyte-specific mARC1 knockdown reverses steatosis and decreases markers of fibrosis in the Gubra DIO-NASH mouse model

To determine the impact of hepatocyte-specific knockdown of Mtarc1 on hepatic steatosis, inflammation, and fibrosis in vivo, a diet-induced obese (DIO)-NASH mouse model was used. Briefly, C57BL/6JRj mice fed a high-fat, high-fructose, high-cholesterol diet for 44 weeks were randomised based on biopsy results after 36 weeks and treated weekly with GalNAc-siMtarc1 or PBS for an additional 8 weeks. GalNAc-siMtarc1 effectively decreased Mtarc1 mRNA and slightly, but significantly, also reduced Mtarc2 mRNA (Fig. 3A and Fig. S5A). GalNAc-siMtarc1 decreased the liver-to-body weight ratio without altering body weight (Fig. 3B and Fig. S5B). GalNAc-siMtarc1 decreased histologically determined steatosis (Fig. 3C and Table 2). In addition, intrahepatic concentrations of TG and TC determined by extraction were decreased by GalNAc-siMtarc1 treatment, whereas plasma TG increased and plasma TC was reduced (Fig. 3D–G). Plasma liver enzymes (Fig. S5C and D) and metrics of hepatic inflammation were unchanged by mARC1 knockdown (Fig. S5E-H and Table 2). GalNAc-siMtarc1 did not alter total fibrosis, as assessed by Sirius Red staining, or levels of collagen 1 but reduced hepatic α-smooth muscle actin (α-SMA) staining (Fig. 4A–C and Table 2). In addition, expression of several genes involved in fibrogenesis were downregulated by GalNAc-siMtarc1 (Fig. 4D). Mild hepatocyte ballooning was observed in the GalNAc-siMtarc1-treated group (Table 2) but as glycogen accumulation can confound the quantification of hepatocyte ballooning,²⁹ measurement of hepatic glycogen was included in a second animal experiment. The majority of the effects of mARC1 knockdown on hepatic phenotypes were reproduced in the second study, including mild hepatocyte ballooning (Fig. S6A-F), and hepatic glycogen was significantly elevated in the GalNAc-siMtarc1 group (Fig. S6G).

Fig. 3.

Hepatocyte-specific Mtarc1 knockdown reverses steatosis in a DIO-NASH mouse model.

C57BL/6JRj mice fed a high-fat, high-fructose, high-cholesterol diet for 44 weeks were randomised based on biopsy results and treated weekly with a GalNAc-siMtarc1 or PBS for 8 weeks. (A) mRNA expression for Mtarc1. (B) % Liver weight of BW. (C) Liver TG as mg/g of liver. (D) Liver TC as mg/g of liver. (E) Plasma TG (mmol/L). (F) Plasma TC (mmol/L). Data are presented as mean ± 95% CI, n = 10–14, ∗p ≤0.033, ∗∗∗p ≤0.001 (unpaired t test). BW, body weight; GalNAc, N-acetylgalactosamine; GalNAc-siMtarc1, GalNAc-conjugated siRNA targeting murine Mtarc1; Mtarc1, mitochondrial amidoxime-reducing component 1; NASH, non-alcoholic steatohepatitis; TC, total cholesterol; TG, triglycerides.

Table 2.

Effect of GalNAc-siMtarc1 on histological scoring of livers.

	Steatosis∗∗∗		Lobular inflammation		Ballooning degeneration∗∗∗		Fibrosis stage
	PBS	GalNAc-siMtarc1	PBS	GalNAc-siMtarc1	PBS	GalNAc-siMtarc1	PBS	GalNAc-siMtarc1
1-point worsening	0.00%	0.00%	0.00%	0.00%	0.00%	78.60%	0.00%	0.00%
No change	100.00%	7.10%	85.70%	57.10%	100.00%	21.40%	71.40%	71.40%
1-point improvement	0.00%	78.60%	14.30%	21.40%	0.00%	0.00%	28.60%	28.60%
2-point improvement	0.00%	14.30%	0.00%	21.40%	0.00%	0.00%	0.00%	0.00%

An unpaired t test was performed for each parameter.

∗∗∗p ≤0.001.

GalNAc, N-acetylgalactosamine; GalNAc-siMtarc1, GalNAc-conjugated siRNA targeting murine Mtarc1; Mtarc1, mitochondrial amidoxime-reducing component 1.

Fig. 4.

Hepatocyte-specific Mtarc1 knockdown decreases markers of fibrosis in a DIO-NASH mouse model.

C57BL/6JRj mice fed a high-fat, high-fructose, high-cholesterol diet for 44 weeks were randomised based on biopsy results and treated weekly with a GalNAc-siMtarc1 or PBS for 8 weeks. (A) Fibrosis as PSR, % area fraction. (B) Col1 as % area fraction. (C) α-SMA as % area fraction. (D) mRNA expression for fibrosis-associated genes; Timp1, TgfB, Col1a1, Mmp2, and Mmp9. Data are presented as mean ± 95% CI, n = 10–14, n.s., ∗p ≤0.033, ∗∗p ≤0.002, ∗∗∗p ≤0.001 (unpaired t test). α-SMA, α-smooth muscle actin; GalNAc, N-acetylgalactosamine; GalNAc-siMtarc1, GalNAc-conjugated siRNA targeting murine Mtarc1; Mtarc1, mitochondrial amidoxime-reducing component 1; NASH, non-alcoholic steatohepatitis; PSR, Picro-Sirius Red.

mARC1 knockdown results in transcriptional changes related to lipid metabolism

To determine the global transcriptional impact of MTARC1 loss of function, RNA sequencing on PHH treated with siMTARC1 or siNT (Supplementary Materials and Methods) was performed. In addition to MTARC1, siMTARC1 altered 17 mRNAs (Fig. 5A and Table S12) enriched in processes and pathways including response to lipid and lipid transport (Table S13). Gene set enrichment analysis identified alterations in lipid and FA metabolic processes and the peroxisome proliferator-activated receptor (PPAR) signalling pathway (Fig. S7A and Table S14). RNA sequencing of the liver of the GalNAc-siMtarc1-treated DIO-NASH mice revealed over 600 differentially expressed protein coding genes (Supplementary Materials and Methods and Table S15) compared with vehicle-treated mice. Although there was no overlap with regulated mRNAs in PHH (except Mtarc1), effects of GalNAc-siMtarc1 on genes involved in the PPAR signalling and lipid metabolism were observed (Fig. S7B, Table S16 and S17).

Fig. 5.

Effect of MTARC1 siRNA in PHH and mouse samples evaluated by omics approaches.

(A) RNA sequencing data showing fold change of significantly dysregulated mRNAs and (B) metabolomic data showing significantly dysregulated metabolites in PHH (green) and mouse tissue from bulk liver prep (purple). FDR ≥0.05 (empty points) or FDR <0.05 (solid points). 3-Hydroxybutyrate (purple star) was significantly reduced in both PHH and mouse samples. (C) Protein-interaction network analysis of significant metabolites revealed significant enrichment in the Kennedy pathway caused primarily by dysregulation of phosphocholine, phosphoethanolomine, and serine (green star). FDR, false discovery rate; MTARC1, mitochondrial amidoxime-reducing component 1; PHH, primary human hepatocytes.

mARC1 regulates the Kennedy pathway and 3-hydroxybutyrate

To gain insight into the metabolic impact of mARC1 knockdown, metabolomic analysis on conditioned media from PHH treated with either siMTARC1 or siNT was performed. Knockdown of mARC1 affected the abundance of 47 metabolites, including increases in O-phosphoethanolamine and phosphocholine, and a decrease in 3-hydroxybutyrate (Fig. 5B and Table S18). A network analysis of putative interacting proteins of the metabolites significantly altered by siMTARC1 identified enrichments in the Kennedy pathway from sphingolipids (WP3933, false discovery rate [FDR] = 0.01) and One-carbon metabolism and related pathways (WP3940, FDR = 0.04) (Fig. 5C). To test if hepatocyte-specific Mtarc1 knockdown was sufficient to impact metabolite concentration in the circulatory system of a whole organism, metabolites in plasma from mice treated with GalNAc-siMtarc1 were profiled. Thirteen of the 47 MTARC1-affected metabolites identified in our PHH samples were also identified in mice (Methods and Table S19). Of these, only 3-hydroxybutyrate was altered in GalNAc-siMtarc1 animals (Fig. 5B and Table S20).

Discussion

Using a combination of genetic approaches and experimental manipulations, a causal relationship between mARC1 expression and liver enzymes, plasma lipids, and liver fat has been demonstrated. Previous genetic findings have been extended^3,8 by establishing novel associations of rare MTARC1 variants with liver enzymes, a shared aetiology of traits at the MTARC1 locus, and a relationship between increased MTARC1 mRNA and worsening of liver enzymes, apo B, and body composition. mARC1 knockdown in PHH decreases lipid content and apo B secretion. In a murine model of NASH, hepatocyte-specific mARC1 knockdown improved hepatic steatosis and decreased markers of fibrosis. Thus, although genetic variation in MTARC1 is causally associated with a range of metabolic traits, hepatocyte-specific decreases in MTARC1 mRNA are sufficient to recapitulate alterations in hepatic and serum lipids.

A common MTARC1 missense variant (p.A165T, rs2642438 G>A) has been associated with all-cause cirrhosis, NAFLD, liver enzymes, and serum lipids.^3,8 Here, HyPrColoc was applied to show that rs2642438 is the likely causal variant across these traits, demonstrating a shared genetic mechanism. In addition to the common missense variant (p.A165T, rs2642438 G>A), we identified novel missense MTARC1 variants associated with liver enzymes. However, as the impact of missense variants in MTARC1 on protein levels, localisation, and activity is unclear, the directional effect of MTARC1 can only be inferred from the rare loss of function variant encoding a premature stop codon (p.R200Ter, rs139321832:C:T).³⁰ Using MR, we demonstrated that increased MTARC1 mRNA was associated with adverse effects on metabolic traits. Although the tissues with available eQTLs (muscle, left heart ventricle, and rectum) may not contribute to the associated phenotypes, the concordant effect direction indicates a consistent impact of the gene variants on gene expression across tissues. Our work reports a novel association of the MTARC1 locus with body composition traits, suggesting an additional metabolic benefit of decreased mARC1 expression, and underscores how genetic variants can have diverse systemic effects likely driven by multiple organs.

The cell type(s) that mediate the effect of genetic variation in MTARC1 on the various metabolic traits has important therapeutic implications. Beneficial effects of hepatocyte-specific MTARC1 inhibition could be therapeutically tractable using GalNAc–siRNA technology. Although PHH are the gold standard model for metabolic modelling, their long-term utility is limited by rapid dedifferentiation in vitro. Using the recently described 5C protocol for long-term PHH culture,²³ we have shown that siRNA knockdown is effective and persistent in this model and lipid accumulation can be stimulated by exposure to an FFA mix and inhibited by knocking down DGAT2, which catalyses the last step of TG synthesis. siMTARC1 decreased lipid accumulation in both the absence and presence of the FFA mixture. The decreased lipid accumulation was not associated with alterations in FA uptake, as assessed by the remaining concentration observed in conditioned media, or oxidation of FA or DNL from glucose. GalNAc-siMtarc1 treatment of Gubra DIO-NASH mice decreased hepatic TG content. PBS was used as a vehicle control rather than a GalNAc-NT siRNA owing to difficulties in assessing proper internalisation without an mRNA knockdown readout and the potential of off-target effects that may confound results. Although a contribution of the GalNAc moiety to the effects of GalNAc-siMtarc1 treatment cannot be excluded, this concern is mitigated by concordant effects of siMTARC1 across model systems with independent siRNA delivery methods. Knockdown of hepatocyte mARC1 decreased neutral lipid content in PHH and mice, suggesting that hepatocyte MTARC1 mediates the genetic associations with NAFLD. This has major therapeutic implications given the clinical utility of GalNAc conjugation to effectively deliver RNAi-based therapies to hepatocytes.

Hepatocytes secrete significant amounts of TG within VLDLs. As the anti-steatotic effect observed in carriers of the A allele of rs2642438 is accompanied by elevated plasma TG, increased VLDL secretion has been proposed as a potential mechanism by which MTARC1 loss of function may reduce hepatic lipid content.⁸ Our findings that hepatocyte mARC1 knockdown increased supernatant and plasma TG in PHH and mice, respectively, support this hypothesis. Furthermore, the decreased secretion of apo B in PHH treated with siMTARC1 parallels decreased apo B in carriers of the A allele of rs2642438.^3,8 This suggests that loss of hepatocyte mARC1 function may decrease the number but increase the size of secreted VLDL particles, leading to reduced apo B and increased plasma TG, respectively. Transcriptomic analysis demonstrated a decreased expression of genes involved in FA metabolic processes and PPAR signalling upon mARC1 knockdown in vitro and in vivo. These findings are consistent with the decreased concentration of 3-hydroxybutyrate in cell culture supernatant and mouse plasma following mARC1 knockdown. Observed alterations in the Kennedy pathway in siMTARC1-treated PHH are consistent with the observation that carriers of the A allele of rs2642438 have increased phosphatidylcholine in liver biopsies, which is essential for the secretion of VLDL.^8,31 Together, these findings suggest that MTARC1 inhibition influences TG secretion, rather than lipid catabolism, to reduce hepatic TG. The molecular mechanisms by which mARC1 exerts its effects on TG metabolism, and whether changes in 3-hydroxybutyrate and the Kennedy pathway are causative, require further study.

Although human genetics links MTARC1 to more advanced forms of NAFLD,[31], [32], [33] hepatocyte-specific knockdown of mARC1 did not improve previously established fibrosis in a DIO-NASH model. This may be attributable to a contribution of non-hepatocyte mARC1 to the genetic signal, species differences, or temporal considerations. The anti-steatotic effect of mARC1 inhibition may be sufficient to prevent, but not reverse, the development of fibrosis in agreement with a Mtarc1-induced reduction of α-SMA. Alternatively, a longer treatment period may be necessary to impact fibrosis in the DIO-NASH model. Future studies will be necessary to determine whether mARC1 alters fibrosis, whether such an effect is mediated via steatosis or direct mechanisms, and when in the disease progression therapeutic intervention may be effective.

In summary, genetically predicted MTARC1 gene expression is associated with multiple metabolic traits, and hepatocyte mARC1 likely mediates the association with hepatic and plasma lipids. Knockdown of mARC1 in PHH and mice decreases intrahepatocellular lipid content and increases TG secretion. This work supports the therapeutic utility of hepatocyte-specific targeting of mARC1 and further studies to elucidate the mechanisms by which mARC1 impacts NAFLD.

Financial support

This work was supported by Novo Nordisk. SRN has a Novo Nordisk Postdoctoral Fellowship run in partnership with the University of Oxford and the British Heart Foundation (Fellowship FS/15/56/31645 and FS/SBSRF/21/31013 to LH).

Authors’ contributions

Performed experiments and procedures: LCL, MAR, LSH, RP, EJL, TND, IS, BG, AC, MEG, AHAE, CED, STH, JN, SRN, EW, SBR, WL. Contributed to concept and design of the study: LCL, MAR, CED, NWP, BA, STH, KC, JF, JMMH, WGH. Contributed to formal analysis and data curation: LCL, MAR, EJL, TND, IS, CED, RRK, EMT, LC, CM, SH, NWP, BA, STH, JF, JN, SRN, EW, SBR, WL. Were involved in original draft preparation: LCL, MAR, CED, RRK, LC, CM, JMMH, BA, STH, SRN. Supervised the study: MAR, BA, JMMH, LH, JF, KC. Reviewed and approved the final manuscript: all authors.

Data availability statement

The authors confirm that the data supporting the findings of this study are available within the article and Supplementary information. The transcriptomic and metabolomic datasets are available in GEO (GSE225616 for mouse and GSE223652 for human) and MetaboLights (MTBLS6909),³⁴ respectively.

Conflicts of interest

A provisional patent application directed to the subject matter disclosed in this manuscript has been filed by Novo Nordisk A/S. LCL, LC, LSH, RRK, CM, JN, CED, STH, RP, IS, EJL, TND, AC, SH, BG, MEG, ETM, AHAE, WGH, KC, JF, JMMH, BA, and MAR are Novo Nordisk A/S or Novo Nordisk Ltd employees. EW and NWP are former Dicerna employees. SBR is a former employee of Dicerna Pharmaceuticals Inc. and presently employed by the Novartis Institutes for BioMedical Research (NIBR), Cambridge, MA, USA. WL is a former employee of Dicerna Pharmaceuticals and presently employed with Biogen Inc, Cambridge, MA, USA. LH is a scientific consultant for Novo Nordisk A/S. The remaining authors declare no competing interests.

Please refer to the accompanying ICMJE disclosure forms for further details.

Supplementary data

The following are the supplementary data to this article: