Divergent role of Mitochondrial Amidoxime Reducing Component 1 (MARC1) in human and mouse

Recent human genome-wide association studies have identified common missense variants in MARC1, p.Ala165Thr and p.Met187Lys, associated with lower hepatic fat, reduction in liver enzymes and protection from most causes of cirrhosis. Using an exome-wide association study we recapitulated earlier MARC1 p.Ala165Thr and p.Met187Lys findings in 540,000 individuals from five ancestry groups. We also discovered novel rare putative loss of function variants in MARC1 with a phenotype similar to MARC1 p.Ala165Thr/p.Met187Lys variants. In vitro studies of recombinant human MARC1 protein revealed Ala165Thr substitution causes protein instability and aberrant localization in hepatic cells, suggesting MARC1 inhibition or deletion may lead to hepatoprotection. Following this hypothesis, we generated Marc1 knockout mice and evaluated the effect of Marc1 deletion on liver phenotype. Unexpectedly, our study found that whole-body Marc1 deficiency in mouse is not protective against hepatic triglyceride accumulation, liver inflammation or fibrosis. In attempts to explain the lack of the observed phenotype, we discovered that Marc1 plays only a minor role in mouse liver while its paralogue Marc2 is the main Marc family enzyme in mice. Our findings highlight the major difference in MARC1 physiological function between human and mouse.


Introduction
Non-alcoholic fatty liver disease (NAFLD) that affects 30% of the US population, is a heterogenous disease characterized by two principal phenotypes: hepatic steatosis, caused by accumulation of excess triglycerides (TGs) in hepatocyte cytosolic lipid droplets and nonalcoholic steatohepatitis (NASH), which is defined by the presence of hepatic steatosis with inflammation, and hepatocellular ballooning, with or without liver fibrosis [1,2].Despite enormous research effort in this area, the molecular mechanisms of NAFLD are not completely elucidated, making it challenging to develop effective NAFLD therapies.To date, several genomeor exon-wide association studies were conducted to identify genetic variants associated with liver fat accumulation, inflammation and fibrosis.These studies revealed several missense variants that are associated with lipid accumulation and NAFLD progression [3][4][5][6][7][8][9][10]. Recent genome-wide association studies discovered that common MARC1 missense variants (p.A165T, rs2642438 G>A and p.M187K, rs17850677 T>A) and several rare MARC1 putative loss of function (pLOF) variants (p.R200Ter, p.R305Ter) in humans are associated with decreased liver fat, lower blood ALT, lower alkaline phosphatase, reduced LDL and total cholesterol, as well as protection from all-cause of liver cirrhosis [11][12][13][14][15][16][17][18][19].MARC1 (mitochondrial amidoxime-reducing component 1, known also as MOSC1 and MTARC1) is a molybdenum containing enzyme anchored on the outer mitochondria membrane.MARC1 forms a complex with cytochrome b 5 type B (CYB5B) and cytochrome b 5 reductase 3 (CYB5R3) and metabolizes a broad range of N-hydroxylated compounds, including amidoximes and N-oxides [20][21][22][23][24].The MARC1 protein crystal structure revealed it has two MOSC (moco sulfurase C-terminal) domains: MOSC_C domain contains a catalytic site with Mo-molybdopterin cofactor (MOCO) in its center; MOSC_N domain function is unknown, but it is proposed to play role in substrate recognition by binding to MOSC_C domain or by regulating MARC1 interaction with its cofactors CYB5B and CYB5R3 [25].Intriguingly, both previously-reported MARC1 missense genetic variants, MARC1 p.A165T and M187K, are located outside of the catalytic domain with the p.A165T variant in an alpha helix of the MOSC_N domain and p.M187K in peptide sequence connecting beta strand and alpha helix in MOSC_N domain [25,26], thus it is unclear how these substitutions contribute to the protective effect against liver disease.
Interestingly, MARC1 closest paralogue MARC2 has a similar protein structure, subcellular distribution and similar, but not identical, preferences to substrates [21,[27][28][29].Whole body Marc2 knockout mouse reported to have reduced body weight, resistance to high fat diet induced obesity and protection from fatty liver [30].Currently, no Marc1 knockout mice have been described in the literature and although recently published study showed liver-specific siRNA Marc1 knockdown led to reduction of hepatic triglyceride accumulation [19], it is unclear how MARC1 pLOF variants affect hepatic metabolism and protect from liver disease in humans.
In this study we used exome-wide association analysis in over 500,000 individuals to identify several novel rare pLOFs variants in MARC1 with phenotypes similar to previously reported p.Ala165Thr (A165T) and p.Met187Lys (M187K) variants [11][12][13][14][15][16][17][18][19].We showed p. A165T substitution in MARC1 causes protein instability, leading to the lower MARC1 protein levels in the livers of human carriers of MARC1 p.A165T variant, confirming this variant is a loss of function.To further understand the mechanism of MARC1 in liver disease, we generated and characterized Marc1 knockout mice, as well as assessed an enzymatic activity of both Marc1 and Marc2 proteins using CRISPR/Cas9 approach in primary hepatocytes.Our studies highlight the major differences between human and mouse MARC1 function and suggest Marc2 is the major Marc family enzyme in mouse liver.

Human MARC1 variant detection
To expand on MARC1 genetics and potentially identify novel MARC1 variants, we performed association testing of genetic variants in the MARC1 gene with liver and other health-related phenotypes in over 540,000 individuals from five ancestry groups who had available exome sequencing data.
In addition to the associations with reduced ALT levels, we found that the burden of rare pLOF variants in MARC1 A165T and M187K were significantly associated with reduced liver fat (Fig 1B), as estimated by proton density of the fat fraction which is quantified using magnetic resonance imaging (PDFF-MRI), and with various liver disease etiologies, including alcoholic and non-alcoholic liver diseases, and liver cirrhosis (Fig 1C and S2 Table ).We next estimated the associations of these MARC1 variants with cardiometabolic traits to better understand the broader impact on health and functional mechanism.We observed significant associations (P<0.001) between MARC1 loss of function variants and lower circulating apolipoprotein B, LDL-C and HDL-C levels, but higher circulating triglyceride levels (S2 Fig and ).In addition, we identified 2 homozygote MARC1 putative loss of function carriers, one individual in UKB, and one in GHS (S1 Table ).Both individuals had no reports of liver disease, had normal BMI, average ALT, blood pressure and lipid levels.

MARC1 genetic variant p.A165T shows protein instability and aberrant protein localization in hepatocytes
First, we sought to evaluate the effect of MARC1 pLOF on MARC1 protein expression.Both the published and our own in silico prediction studies showed that MARC1 p.A165T can cause protein instability as indicated residue is highly conserved and disrupts the alpha helix [26].To assess A165T/M187K MARC1 variant effect on expression and localization of MARC1 protein in vitro, we overexpressed the human wild-type and two MARC1 variants (A165T and M187K) in a human HuH-7 hepatoma cell line.Levels of MARC1 mRNA were comparable in cells expressing wild-type and the two variants of human MARC1, however the level of A165T

PLOS GENETICS
variant protein was reduced by 64%.Interestingly, MARC1 M187K variant protein levels were similar to MARC1 WT (Fig 2A -2C).Identical experiments with MARC1 variant overexpression were repeated on the second human hepatoma cell line (HepG2) with comparable results (S4 Fig) .To avoid a potential non-physiological effect of transfection on protein expression level, we performed human liver protein analysis utilizing livers from A165T variant carriers.Consistent with findings from transient recombinant protein overexpression, the analysis of these human livers revealed 60% reduction in MARC1 p.165T variant protein level compared to the most common (p.165A) variant.Concurrently, the mRNA levels of both variants were similar.No changes were observed in MARC2 mRNA or protein levels in these same samples (Fig 2D -2F).To uncover the cause of the discrepancy between mRNA and the protein level for A165T variant, we performed confocal microscopy studies in HuH-7 cells.These studies revealed that the protein from cells that overexpressed human recombinant A165T variant showed only partial colocalization within mitochondria, while both proteins from cells overexpressing MARC1 wild-type and M187K variants fully colocalize within mitochondria (Figs 3  and S5).These data suggest that human MARC1 substitution of alanine with threonine in A165T led to partially unstable mislocated protein.To evaluate if A165T protein was degraded by proteasomal or lysosomal degradation, we overexpressed wild-type and A165T variant in HuH-7 cells and incubated cells with or without proteasomal inhibitor MG132 and lysosomal inhibitor chloroquine.The use of either inhibitor did not rescue MARC1 A165T variant protein, (S6A and S6B Fig) .In order to differentiate between the compromised translation and stability of the MARC1 protein, we overexpressed wild-type and A165T variant in HuH-7 cells and then treated the cells with the protein synthesis inhibitor cycloheximide (CHX).We also added either MG132 or chloroquine to CHX and compared the expressions of the MARC1 proteins at 0, 3, 6, and 9 hours post-treatment.The addition of MG-132 or chloroquine to CHX for the cells expressing MARC1 p.165T variant did not lead to p.165T variant protein protection from degradation compared to CHX alone treatment, implying that p.165T variant protein instability is not caused by either proteasomal or lysosomal degradation (S6C and S6D Fig).

Marc1 is predominantly expressed in the liver in mice
In humans, MARC1 is mainly expressed in white adipose tissue, thyroid, breast and liver, with adipose expression being 3 times higher than the liver (GTEX portal).To estimate tissue distribution and expression levels of Marc1 mRNA and protein in mouse, we analyzed various tissues from wild type C57BL/6 male mice.Marc1 mRNA was expressed almost exclusively in the liver, with very minor expression in brown adipose tissues (BAT), as well as in the descending aorta and white adipose tissues (Fig 4A).Similarly, Marc1 protein was mainly expressed in the liver, followed by brown and white adipose tissues (Fig 4C).The closest Marc1 paralogous protein Marc2, was highly expressed in all three parts of the small intestine, liver and kidney consistent with its mRNA expression pattern in mice (Fig 4B and 4C), and with liver being the tissue where human MARC2 is expressed at the highest level (GTEX portal).Notably, the tissue expression of two cofactors of Marc1 and 2 enzymes (Cyb5b and Cyb5r3) show an expression pattern in mice similar to Marc2 but not Marc1 (Fig 4C).

Development and validation of Marc1 knockout mice
To support a hypothesis that A165T of MARC1 variant is a loss of protein function mutation and to understand if MARC1 deletion or inhibition may have a beneficial effect on liver disease in humans, we generated whole body Marc1 knockout mouse.The mouse Marc1 gene

Marc1 ablation does not elicit hepatoprotective effect in NASH mouse models
Initial phenotyping of Marc1 KO male mice fed ad lib chow diet showed reduction in liver triglycerides and phospholipids, however no changes in serum liver injury markers (ALT and AST), or other markers were observed (S5 Table ).Liver gene expression profiling of Marc1 KO chow fed mice revealed no major changes in gene expression in the liver, including de novo lipogenesis or oxidation pathways genes.(S7 and S8 Tables).
To determine Marc1 effect on liver lipid accumulation, inflammation, and fibrosis, we challenged Marc1 KO and WT male mice with NASH/NAFLD-causing diets: high fat/high fructose diet (HFHFD) and choline-deficient L-amino acid defined high fat diets (CDAA-HFD).Feeding Marc1 KO and WT mice with HFHFD for 35 weeks did not affect body weight gain, body composition, liver enzymes, and had no effect on serum and liver lipids (Fig 6A-6D and S5 Table ).Similarly, there were no significant differences in liver inflammation or fibrosis gene markers between Marc1 KO and WT mice measured by selected gene expression in the liver by qRT-PCR (Fig 6E).Histological analysis of liver sections from HFHFD fed animals showed no differences between Marc1 KO and WT male mice in lipid droplet size and distribution (hematoxylin/eosin staining, Fig 7A and 7B).Similarly, we observed no differences between genotypes when quantified hepatic inflammation (CD45 stain) and fibrosis (Sirius red stain) (Fig 7A , 7C and 7D).Feeding Marc1 KO and WT male mice with the second NASH-causing diet, CDAA-HFD, for 18 weeks did not reveal any significant variation in body weight, body composition or liver lipid accumulation, hepatic inflammation, fibrosis, or any changes in serum lipid levels (S7A-S7E, S8A-S8D Figs and S5 Table ).To expand our phenotyping to female mice we utilized the same diets (HFHFD and CDAA-HFD), however found similar lack of phenotype in Marc1 KO females: no differences in body weight gain, body composition, liver and serum lipids or liver enzymes between female KO and WT mice (S9 Fig and S6 Table).
Focusing only on the effect of Marc1 deletion on liver steatosis, we fed Marc1 KO and WT male mice with high sucrose diet (HSD, 9 weeks) or high fat diet (HFD, 11 weeks).However, although we observed significant reduction in ALT levels in Marc1 KO vs WT in both diets, we did not detect any significant difference in liver triglycerides between WT and Marc1 KO mice on these diets, consistent with the findings from HFHFD and CDAA-HFD studies described above (S5 Table ).

Marc2 but not Marc1 is a main enzyme with N-hydroxylated amidine reduction properties in mouse liver
To assess Marc1 and 2 enzymatic activities for reduction of N-hydroxyl substrates, we adapted an in vitro system utilizing a commonly used synthetic substrate, benzamidoxime (BAO), to evaluate its reduction to benzamidine (BA) [28].Our initial attempts to use human hepatic or other cell lines (HepG2, ZR-75-1 and HEK293) with knockout or overexpression of human MARC1 gene described earlier [21,27,28,31] did not show any reliable and consistent measurable BAO reduction and BA production.Thus, we decided to utilize freshly prepared primary mouse hepatocytes to assess N-hydroxyl substrate reduction activity by Marc1 and Marc2 enzymes directly.Our early experiments with plated wild type mouse primary hepatocytes showed BAO reduction was detectable in 2-4 hours' time period, reaching up to 30% conversion of BAO to BA in the media.However, no difference was observed in BAO reduction rates between primary hepatocytes isolated from Marc1 KO or WT mice.Thus, we hypothesized that the conversion of BAO to BA observed in primary hepatocytes may be driven by Marc1 paralogue, Marc2, which is present in both Marc1 KO and WT hepatocytes.To evaluate this  Marc1/2 gene editing (inactivation) as CRISPR/Cas9 editing happens independently in each hepatocyte; we cannot also exclude that mouse hepatocytes may possess other enzymes with low affinity for BAO reduction.It is worth noting that CRISPR/Cas9 editing could have offtarget effect even after careful gRNA selection [32].

Marc2 knockout mouse develop neurological defects
Marc1 and Marc2 proteins are highly similar (59%) and share the same substrate preference as mentioned earlier.While our human genetic analysis did not suggest MARC2 plays a role in liver disease in humans (S3 Fig) , we decided to generate Marc2 KO mice in the attempt to utilize Marc2 deletion as a surrogate that can help to shed light on the role of molybdenum-containing enzymes on liver disease-relevant conditions in mice.Marc2 KO mice were generated on 100% C57BL/6 genetic background, using the same technique of gene editing as was used 12 weeks old male Cas9 transgenic mice were injected in tail vein with 3x10 10 genome copies (GC) of AAV8 containing Cas9 gRNA against Marc1, Marc2, Marc1/2 (edit both genes), and control (red fluorescent protein (RGF)).(A) 3 weeks after virus injection mice were sacrificed and primary hepatocytes isolated, counted and plated on poly-D-lysine pre-coated plates, then cells were left in incubator to attach to plate for 4 hours.After attachment, cells were washed with PBS once and incubated in serum free media with 0.5 mM benzamidoxime for 2 hours, then cell media were collected and protein from all primary cell plates was extracted, size fractioned on 4-20% SDS-PAGE gel and blotted with corresponding antibodies.Densitometry was used to evaluate protein reduction after Marc1 or Marc2 gene editing and normalized against house-keeping gene (calnexin).(B) BAO and its metabolite benzamidine (BA) were extracted from the collected cell media and measured by UV-HPLC.Controls: no cells-no primary hepatocytes plated for incubation of 0.5 mM BAO in serum free media for 2 hours (control for spontaneous conversion from BAO to BA); no BAO-no benzamidoxime added to RFP gRNA treated cells with serum free media (control for BAO/BA in media or cells).Experiment repeated once with comparable results.Red arrows-% reduction of BAO conversion. https://doi.org/10.1371/journal.pgen.1011179.g008 for Marc1 KO mouse.Unexpectedly, at the age of 7-10 weeks 100% of both sexes of homozygous Marc2 knockout mice showed rapidly progressing hind limb paralysis.Movement disabilities significantly affected the possibility of the future phenotyping of these mice on NAFLD/NASH-causing diets.Limited characterization of the young mice on chow diet showed reduced body weights and decreased fat mass in Marc2 KO female mice compared to wild-type (S10 Fig) , consistent with the previously reported findings [30].Marc2 KO mouse liver RNAseq profiling showed a very significant increase of xenobiotic-metabolizing enzymes (such as Fmo3, Cyb2b9, and others), which is consistent with its reported role in detoxification pathways [33]; this signature was not present in Marc1 KO mouse (S8 Table ).The cause of Marc2 KO neurological phenotype is unknown at present as our human genetics analysis did not reveal any association between human MARC2 pLOFs alleles carriers and neurological defects.

Discussion
Recent advances in human genetics have led to identification of several genes of interest that may serve as potential therapeutic targets for non-alcoholic liver disease.MARC1 represents one of them, where newly published loss-of-function mutations in MARC1 were shown to be protective against liver diseases, including most causes of liver cirrhosis [11,17].In this study, we replicated and expanded on MARC1 genetic findings by conducting whole exome sequencing analysis in over 540,000 individuals from five ancestry groups.Our study identified several novel rare, putative loss of function variants in MARC1 gene and showed that these variants have a phenotype similar to the previously reported common variant p.A165T and are associated with protection from liver disease.Additionally, we found that the most common MARC1 gene variant, p.A165T, causes loss of function of MARC1 protein due to its aberrant subcellular localization and degradation.We generated whole body Marc1 knockout mice in order to understand the mechanism of the potential hepatoprotective effect of MARC1 deletion.However, our study found that whole-body Marc1 deficiency in mouse is not protective against hepatic triglyceride accumulation, inflammation or fibrosis under different dietary challenges that are known to induce these conditions.In our attempts to explain the lack of observed phenotype that contradicts the findings from human genetic studies, we discovered Marc1 plays only a minor role in mouse liver, while its paralogue, Marc2, is the main Marc family enzyme in mice, once again stressing the differences between human and mouse biology.
Mechanistically, it is unknown how MARC1 pLOFs affects liver disease in humans.Previous studies, using recombinant proteins isolated from E. coli, reported A165T variant does not cause difference in N-reduction of benzamidoxime compared to p.165A variant [33,34].Our current assessment of the most common MARC1 p.A165T variant in hepatic cells revealed that it produces partially mislocalized and unstable protein compared to the MARC1 wild type.This finding confirms that MARC1 p.A165T variant is a loss of function.Previous human MARC1 protein crystal structure studies indicated that MARC1 p.A165T variant is located in the conserved MOSC_N domain alpha helix, which lies far away from the enzyme catalytic site in MOSC_C domain [25].Thus, the A165T substitution is unlikely to affect the catalytic activity of the enzyme directly, but rather diminishes its activity through reduction in hepatic expression due to the disturbed protein intracellular localization and instability.Inconsistently with this finding, the results from overexpression of another MARC1 pLOF variant p. M187K in liver cell lines did not show changes in protein level or its colocalization within mitochondria (Figs 2A, 2B, 3, S4 and S5).While this result is puzzling and may suggest A165T and M187K affect the enzymatic activity of MARC1 differently, it was challenging to assess due to the luck of reliable in vitro or in vivo systems.We can only conclude that overexpression of M187K variant in cell lines may not fully reproduce physiological conditions.Unfortunately, due to rareness of M187K variant, it was not feasible to confirm its mRNA and protein levels in human liver as it was performed for A165T variant (Fig 2D and 2E).
While multiple genes identified through human genetic analysis shown strong association with liver disease, their mechanistic studies in mouse models proven to be challenging.Current NASH/NAFLD mouse models are far from an ideal instrument for these studies due to disease progression and pathology being different from humans, differences in tissue-specific expression between mice and humans, as well as potential lack of essential co-factors or intermediates.Thus, knockout mice of recently discovered protein-truncating variant in HSD17B13, that leads to generation of unstable protein (essentially knockout) and associated with protection from liver disease in humans, did not show any protective phenotype in mice, or even led to unexpected worsening of the phenotype [35,36].Original studies in Pnpla3 knockout mice of genetically validated high risk-NAFLD associated allele I148M, led to enzymatic loss of function of PNPLA3 protein, but showed no phenotype in mouse [37,38].The follow-up studies discovered that I148M substitution leads to excessive accumulation of the protein on lipid droplets-a new, modified function of the protein leading to the major phenotype difference between KO (lack of phenotype) and I148M knock-in mice (increase in fat accumulation in the liver) [39].It is noteworthy, that even though I148M has a fat-accumulation phenotype in mouse liver, this phenotype is minor compared to the one observed in humans.This discrepancy is driven by multiple factors with one of them being the difference in mouse vs human expression of PNPLA3: the expression of PNPLA3 is high in human liver but low in adipose tissue, while contrarily it is very low in mouse liver and high in mouse adipose tissue [40,41].
Nevertheless, in the effort to understand the mechanism of how MARC1 pLOF leads to protection from liver disease, we generated Marc1 knockout mice and challenged them using the most validated diets, HFHFD and CDAA-HFD, that characterized by the development of all three major stages of NAFLD: steatosis, inflammation and fibrosis.Our attempt to understand how reduction in Marc1 enzyme can lead to protection from liver disease had some limitations.Specifically, MARC1 mRNA tissue-specific expression comparison between mouse and human revealed that in mouse, Marc1 mRNA is mostly expressed in liver, followed by very minor expression in brown and white adipose tissues (Fig 4 ), where human MARC1 mRNA is expressed at highest levels in adipose, with liver expression almost three-fold lower than in adipose tissue (GTEX Portal).Thus, liver is not the main organ expressing MARC1 in humans, and it could not be excluded that MARC1 pLOF hepatoprotective effect is mediated by adipose tissues, and not by the liver.Intriguingly, a recently published study utilizing siRNA for reduction of Marc1 expression in mouse liver suggested that liver-specific deletion of Marc1 is sufficient to see the hepatoprotective phenotype in rodents, however this study did not utilize nonspecific siRNA control [19].Our data does not support those findings, as we did not observe any hepatoprotective phenotype in both sexes of whole body homozygous Marc1 KO mice using multiple NASH/NAFLD diets (S5 and S6 Tables).
Conspicuously, our studies in mouse primary hepatocytes using CRISPR/Cas9 deletion of Marc1, Marc2 or Marc1/2 genes showed that in the mouse liver, Marc2 is the main enzyme responsible for N-reductive biotransformation of N-hydroxylated substrates, which is congruent with the reported data showing major reduction in N-reductive activity in Marc2 KO mouse liver lysates [30].Unfortunately, we have not been able to evaluate the effect of Marc2 deletion on a liver disease-relevant phenotype, due to the severe neurological defect observed in these mice.Mild neurological symptoms (decreased prepulse inhibition and increased startle reflex) were also recently described by European Conditional Mouse Mutagenesis Program (EUCOMM/IMPC) (https://www.mousephenotype.org/data/genes/MGI:1914497).The overall phenotype for this mouse was consistent with a previous report by Clement's group [30]; however Clement's paper does not report any neurological findings.The discrepancy between our data and the published studies is unclear, but the differences could be driven by alternative genetic deletion strategies of Marc2 gene, as well as differences in diet, housing condition or other factors.
In conclusion, our study identified several novel pLOFs in MARC1 that are associated with reduced liver fat and liver enzymes, consistent with previously reported common variant p. A165T.Our work revealed MARC1 A165T substitution leads to protein degradation and mislocalization, confirming that pLOF A165T is a genuine loss of protein function in humans.While our studies pinpoint Marc1 paralogue Marc2 as a main molybdenum-containing enzyme with N-hydroxylated amidine compounds reducing properties in mouse liver, the lack of a phenotype in Marc1 KO mice highlights the significant differences in human and mouse MARC1 protein functions, as well as shows challenges in transition from human genetic findings to deeper mechanistic studies of MARC1 protein in animal models.

Ethics statement
Ethical approval for UK Biobank (UKB) was provided by the North West Multi-center Research Ethics Committee (MREC; 11/NW/0382), which covers the UK.The GHS MyCode study (GHS) is a health system-based cohort of patients from Central and Eastern Pennsylvania (USA) recruited in 2007-2019.GHS was approved by the Geisinger Institutional Review Board (2006-0258).All used cohorts patients/participants provided their written informed consent to participate in this study.

DNA sequencing and genotyping data
High-coverage whole-exome sequencing was performed using NimbleGen VCRome probes or a modified version of the xGen design from IDT.Sequencing was done using Illumina v4 HiSeq 2500 or NovaSeq instruments, achieving over 20x coverage for 96% of VCRome samples and 99% of IDT samples.Variants were annotated using snpEff and Ensembl v85 gene definitions, prioritizing protein-coding transcripts based on functional impact.pLOF and missense variants were classified for functional impact according to in silico prediction algorithms.Common variant genotyping was performed on one of four SNP array types: Illumina OmniExpress Exome array (OMNI; 59345 samples from GHS), Illumina Global Screening Array (GSA; UPENN-PMBB, BioMe and 82,527 samples from GHS), Applied Biosystems UK BiLEVE Axiom Array (49,950 samples from UK Biobank (UKB)) or Applied Biosystems UK Biobank Axiom Array (438,427 samples from UKB).We retained variants with a minor allele frequency (MAF) >1%, <10% missingness, Hardy-Weinberg equilibrium test P-value>10 −15 .We imputed the genotyped variants using the TOPMed reference panel [42] using the TOPMed imputation server [43].In GHS, imputation was performed separately by genotyping platform.Further details are provided elsewhere [10].
Genetic analyses were stratified by ancestry and adjusted for age, age2, sex, age-by-sex and age2-by-sex interaction terms, experimental batch-related covariates, the first 10 common variant-derived principal components, the first 20 rare variant-derived principal components using REGENIE.Results were combined across subsets by fixed-effect inverse varianceweighted meta-analysis [10].

Human phenotype definitions
The data cleaning process for continuous traits involved removing non-physiologic lab values and results stemming from invalid or contaminated specimens.The median transaminase values for each individual were extracted from electronic health records at Geisinger Health System (GHS).In the UK Biobank (UKB), transaminases were measured using a Beckman Coulter AU5800 at the baseline study visit.Cases of binary diseases were defined based on one or more of the following criteria: self-reported disease obtained from digital questionnaire or interview with a trained nurse, in-patient hospitalization for the disease, or clinical-problem list entries of the disease, according to International classification of Diseases, Ninth (ICD-9) or Tenth (ICD-10) Revision diagnosis code, medical procedures, or surgery due to the disease, death due to the disease, and a disease diagnosis code entered for two or more outpatient visits in separate calendar days.Controls were defined as individuals who did not meet any of the criteria for case status.To minimize misclassification, the liver disease control group was excluded from analysis if (i) non-cases were diagnosed with any type of liver disease (not restricted to the type of liver disease in question), (ii) non-cases with only one out-patient encounter related to the type of liver disease in question, (iii) non-cases diagnosed with ascites presumably related to liver failure, and (iv) non-cases with elevated alanine aminotransferase (ALT) levels (>33 IU/L for men, >25 IU/L for women).For more details on the specific entries and population characteristics, please refer to [10].
Marc2 knockout mice (ID #20646) were generated similarly as Marc1 KO mouse using Regeneron VelociGene technology, where second exon of Marc2 gene was replaced with LacZ reporter.Marc2 mice were bred on 100% C57BL/6NTac background.
Male and female mice (one to five per cage) were maintained on a 12 h light/dark cycle at 22 ± 1˚C and were fed ad libitum chow (LabDiet, Cat.No 5053).In some experiments, mice were fed high fat high fructose diet (Research Diets, Cat.No D09100310), choline-deficient Lamino-defined (CDAA) diet (Research Diets, Cat.No A16092201i), high fat diet (Research Diets, Cat.No D12492; 60% fat by calories) or high sucrose diet (Research Diets Cat.No D19022101), as indicated in figure legends.Mouse experiments were performed on age-, sexand strain-matched littermates.All animal procedures were conducted in compliance with protocols approved by the Regeneron Pharmaceuticals Institutional Animal Care and Use Committee.

Mice and human genotyping
DNA from mouse and human livers were extracted by DNeasy Blood & Tissue Kit (Qiagen).DNA area including mouse Marc1 Exon 2 and 3 regions was amplified using Platinum SuperFi II DNA Polymerase (Thermo Fisher) using primers 5 0 -aattgctgctacctggtgct-3 0 and 5 0 -tggttcatgagggttgtcgg-3 0 .PCR product was digested by EcoRV restriction enzyme (New England Biolabs) and separated on 0.8% Agarose gel by electrophoresis.Pair of primers 5 0 -gctaggagcagcttttctga-3 0 and 5 0 -caacagagccgaggtcatca-3 0 were used to amplify whole human MARC1 gene exon 3 by PCR reaction, and ExoSAP-IT Express PCR Product Cleanup Kit (Thermo Scientific) was used to remove both primers before sequencing.Human MARC1 165A/T and 187M/K genetic variants were detected by Sanger sequencing.

Human samples
Fresh frozen human liver samples were acquired from various vendors (LifeNet Health, BioIVT and QPS Medicals).Samples were selected by shortest time period between patient deceased and when liver samples were collected, no RNA degradation (RNA integrity number (RIN) score), absence of hepatitis B and C, liver oncology, severe liver inflammation/fibrosis and minimal or no alcohol use history (all reported by vendor).

RNA preparation and RNA-sequencing read mapping
Total RNA was purified from Wild type and heterozygous/homozygous Marc1 or Marc2 knockout mice tissues using MagMAX-96 for Microarrays Total RNA Isolation kit (Life Technologies), according to manufacturer's specifications.Genomic DNA was removed using Mag-MAX Turbo DNase buffer and TURBO DNase from the MagMAX kit listed above.Messenger RNA was purified from total RNA using Dynabeads mRNA purification kit (Invitrogen).Strand-specific RNA sequencing (RNA-seq) libraries were prepared using KAPA mRNA-Seq Library Preparation kit (Kapa Biosystems).After mRNA quality control analysis, RNA-seq libraries were prepared and long read sequencing (75bp, single end, 15M reads) was used for all samples.Sequencing was performed on Illumina HiSeq 2500 by a multiplexed single-read run with 33 cycles.Raw sequence data (BCL files) were converted to FASTQ format via Illumina bcl2fastq v2.17.Reads were decoded based on their barcodes and read quality was evaluated with FastQC (www.bioinformatics.babraham.ac.uk/projects/fastqc/).Reads were mapped to the mouse genome (NCBI GRCm38) using ArrayStudio software (OmicSoft Corp.) allowing two mismatches.Differential expression analysis was performed using DESeq2 (v1.38.2) with default parameters.

qRT-PCR (Taqman)
Liver and white adipose tissue (~50 mg) or cell line (HuH-7) pellet were homogenized in Trizol (Thermo Fisher), and chloroform was used for phase separation.The aqueous phase, containing total RNA, was purified using MagMAX-96 for Microarrays Total RNA Isolation Kit (Life Technologies), according to manufacturer's instructions.Genomic DNA was removed using RNase-Free DNase Set.SuperScript VILO Master Mix (Thermo Fisher) was then used to reverse-transcribe mRNA into cDNA.The resultant cDNA was amplified with SensiFAST Probe Lo-ROX using the 12K Flex System (BioCat.).Data are presented as a change in gene expression relative to the housekeeping gene, mouse beta-actin (Actb) or human 36B4 (RPLP0), determined using the delta-delta-Ct method [46].RT-PCR (Taqman) primer sequences for mouse Marc1 mRNA detection: probe sequence CCTCCAGTGCAGAGTG-CATGGC, forward primer: GCCTGCCACAAACCCACT and reverse primer PLOS GENETICS GAGCTGCATCCTCTCCACAATC, targeting mouse Marc1 mRNA exon 2 and 3 of open reading frame.

Expression of MARC1 in cultured hepatocytes
The cDNA for human MARC1 (NCBI NM_022746.4)was cloned downstream of the constitutive HSV thymidine kinase (pTK) [47] or CMV promoter in pcDNA 3.1+ plasmid backbone.Single or triple FLAG tags (DYKDDDDK) with GGGGS linker were placed at the C-terminus of each plasmid construct.In some initial studies pCMV-XL6 (C-terminal 1x FLAG/MYC tag, CMV promoter) plasmids (Origene) were used [5].Single nucleotide changes were introduced using the QuikChange site-directed mutagenesis kit (Stratagene) and confirmed by Sanger sequencing.

Assessment of MARC1 protein degradation
Plasmids encoding empty vector (EV), wild-type (WT) and common variants (165T) of human MARC1 (with C terminal 1x FLAG tag, pcDNA 3.1+ plasmid backbone) were transfected in HuH-7 cells using TransIT-LT1 Transfection Reagent (Mirus, Cat.No MIR 2300) according to manufacture protocol.Cells were grown 48 hours in full media (high glucose DMEM with 10% full calf serum (FCS) plus 100 IU/ml penicillin and 100 μg/ml streptomycin), then DMSO only (vehicle) or 10 uM MG-132 in DMSO (proteasomal inhibitor) or 10ug/ml chloroquine in water (lysosomal inhibitor) plus DMSO separately in the same dish were added in cell media, cell incubated for additional 8 hours.Later, cells were collected by scraping, solubilized in 1x RIPA buffer (Millipore) with protease inhibitors, proteins measured, denatured/ reduced, and separated on SDS-PAGE (4-20%) gels and blotted with corresponding antibodies Flag tag (MARC1-Flag), LC3I/II, Ubiquitin and Calnexin antibodies.Ubiquitin was used as proteasomal inhibitor control, LC3B-I/II as lysosomal inhibitor control [48].In the cycloheximide (CHX) chase assay, CHX (Sigma-Aldrich, Cat.No C7698; 300 μg/ml in DMSO) was added dropwise to the full media of HuH-7 cells.The degradation of the MARC1 protein was then assessed at 0, 3, 6, and 9 hour intervals.In addition, a combination treatment of CHX (300 μg/ml) plus MG-132 (10 uM) or CHX (300 μg/ml) plus chloroquine (10ug/ml) was utilized.DMSO was used for the control group treatment and was added to each well with the combinational treatments.

Marc1 antibody production and validation
Polyclonal anti-mouse Marc1 antibodies were produced by immunization of rabbits with fragment of mouse Marc1 protein (NCBI NP_001277202.1,aa191-340) produced in E. coli.After the last immunization of rabbits, serum antibodies were affinity purified using mouse Marc1 specific peptide CSEQALYGKLPIFGQYFALENPGTI (aa 306-329; GenScript).Generated antibodies were validated on wild type and Marc1 KO mouse liver samples using Western Blot.In most experiments purified rabbit anti mouse Marc1 IgG (batch DB0161 (1ug/mL)) was used.

Liver histology
Liver tissues were collected from isoflurane-anesthetized/cervical dislocated mouse, fixed in 4% paraformaldehyde (PFA) for 24 hours.The fixed tissues were embedded in paraffin, sectioned, and stained with hematoxylin and eosin (Histoserv, Inc.).
Picro-Sirius Red staining was done on PFA fixed, unstained paraffin embedded liver sample slides.After heating, the slides were deparaffinized with xylene, ethanol, and distilled water, then stained in Bouin's solution (Sigma Aldrich Cat.No HT10132) and 0.1% Picro-Sirius Red solution (Sigma Aldrich Cat.No 365548), fixed with 0.5% acetic acid solution.Finally, slides were washed and mounted using Cytoseal (Epredia, Cat.No 831016).
For CD45 immunohistochemistry staining, automated staining was performed on Leica Bond RX autostainer (Leica Biosystems).Slides were deparaffinized with Bond Dewax solution (Leica Biosystems, Cat.No AR9222) followed by antigen retrieval with ER2 (EDTA-based) buffer (Leica Biosystems, Cat.No AR9640) at 100˚C for 30 minutes.Anti-CD45 antibody (Cell Signaling, Cat.No 70257) was added at 0.1 ug/ml followed by 1 hour incubation at room temperature, followed by chromogenic detection with Bond Polymer Refine Detection kit (Leica Biosystems, Cat.No DS9800) that includes anti-rabbit HRP polymer, DAB and hematoxylin counterstain.Slides were washed between steps with Bond wash solution (Leica Biosystems, Cat.No AR9590).After staining, slides were dehydrated in increasing concentrations of ethyl alcohol and xylenes and mounted with Surgipath Micromount mounting medium (Leica Biosystems, Cat.No 3801731), dried and scanned on Aperio AT2 scanners (Leica Biosystems).
To measure area of cytosolic lipid droplets, fibrosis or CD45 positive cells, stained slides were scanned with Leica Aperio AT-2 Scanner and whole tissue area, non-stained area, and the area of vessels were qualified by HALO Classifiers software.The area of lipid droplets/ fibrosis/CD45 (%) was calculated by subtracting the area of vessels from non-stained area and then divided by the whole tissue area.

Imaging of MARC1 intracellular localization
HuH-7 cells (Creative Bioarray, Cat.No CSC-C9441L) were cultured in complete medium (high glucose Dulbecco's modified Eagle's medium with 10% fetal calf serum (FCS) plus 100 IU/ mL penicillin and 100 mg/mL streptomycin.On day 0, cells were transfected with pcDNA 3.1+ (CMV promoter) containing plasmids with no gene (empty vector) or human MARC1 wild type and common variants with C terminal FLAG/MYC tags.Transfection was done with Neon Transfection System (Thermo Fisher, Cat.No MPK5000/MPK1025) with pre-validated settings (pulse-1230 V; pulse width-20 ms; pulse numbers: 3) according to producer instructions.Cells were plated on μ-Slide dishes covered with ibiTreat (Ibidi, Cat.No 80826) in full media for 48 hours, after cell media was changed to pre-warmed serum-free media with 200nM Mitotracker Orange CMTMRos staining solution (Thermo Fisher, Cat.No M7510) and incubated for 20 minutes in 37˚C 5% CO 2 in the dark.Cells were washed, fixed and permeabilized with Triton X100 and blocked with 5% donkey serum supplemented by 2% BSA.After the final wash, cells were incubated with Flag antibody (Sigma Aldrich, Cat.No F1804; 1:1000) for 1h RT, washed and incubated with secondary antibody (Alexa Fluor 647 AffiniPure Donkey Anti-Mouse IgG (Jackson ImmunoResearch Inc., Cat.No 715-605-151) RT for 1 hour.After the final wash, cells were incubated in Mounting Medium with DAPI (Ibidi, Cat.No 50011) and imaged with Leica confocal microscope.
Marc1/Marc2 benzamidoxime (BAO) reduction assay in mouse primary hepatocytes 10-12 week old male Cas9 transgenic mice (Cas9-Ready V2.5) were injected in tail vein with 3x10 10 genome copies (GC) of AAV8 containing Cas9 gRNA against Marc1, Marc2, Marc1/2 (one gRNA targeting both genes) or control gRNA (anti red fluorescent protein (RGF)).2-4 weeks after virus injection, mice were sacrificed by lethal isoflurane inhalation and cervical dislocation, and primary hepatocytes isolated, counted and plated on poly-D-lysine pre-coated plates [49].Cells were kept in incubator (37˚C, 5% CO 2 ) in full media (Williams E media with 10% full calf serum (FCS) plus 100 IU/ml penicillin, 100 μg/ml streptomycin and 2 mM L-glutamine) to attach to the plate (4h).After attachment, cells were washed with PBS once and incubated in serum-free media (low glucose DMEM plus 100 IU/ml penicillin, 100 μg/ml streptomycin and 2 mM L-glutamine, 10 μM sodium molybdate dihydrate) with 0.5 mM benzamidoxime for 2 hours.After reaction, cell media was collected, spun 1000 x G for 5 min at 4˚C, supernatant collected, reaction stopped by adding ultra-pure methanol (1:1, vol/vol), spun 21000 x G 5 min 4˚C, supernatant collected and diluted in 20 mM ammonium bicarbonate 1:5 (vol/vol).After media collection, protein from all primary cell plates were extracted by 1x RIPA buffer with protease inhibitors, total protein amount measured by BCA assay.Reduced protein lysates in Laemmli SDS sample buffer were size-fractionated on 4-20% SDS-PAGE gel and blotted with corresponding antibodies.
Benzamidoxime (BAO) and its metabolite benzamidine (BA) were measured in cell media by HPLC-UV.All HPLC analyses were performed using a Waters ACQUITY I-Class UPLC system Separation Module with a Waters ACQUITY UPLC Photodiode Array Detector.Waters Empower 3 Build 3471 integration software was used for data analysis.Benzamidoxime and benzamidine were separated on a Waters Viridis CSH Fluoro-Phenyl Column with a particle size of 1.7 μm, a diameter of 2.1 mm, and a length of 50 mm.The mobile phase was composed of 20 mM ammonium bicarbonate at pH 8.5-8.6 and 30% acetonitrile (vol/vol) under isocratic conditions with a flow rate of 0.2 mL/min and a sample injection volume of 2 μL.The total run time was 5 minutes (retention time of benzamidoxime: 1.1 ± 0.2 minute, benzamidine: 1.7±0.2minute), and the detection wavelength was set at 229 nm.The column temperature was maintained at 40˚C, while the sample storage temperature was kept at 5˚C.To quantify Benzamidoxime depletion and Benzamidine conversion, calibration curves were used with concentration ranges of 0.1-2 mM and 0.00832-1.04mM, respectively.In addition, a diluted cell media blank was injected to check for any potential interference in both benzamidoxime and benzamidine measurements.All standards were prepared in experiment used cell media to closely mimic the sample matrix.

Statistical analysis
Statistical and graphical data analyses were performed using Microsoft Excel and Prism 9 (GraphPad Software, Inc.).Mean values were compared using unpaired two-tailed t-tests or one-way ANOVA.Data shown as mean values with standard error.In box and whisker plots, the middle line is plotted at the median, the upper and lower hinges correspond to the first and third quartiles, and the upper and lower whiskers display the full range of variation (minimum to maximum).Grubbs' test was used to determine and remove significant outliers.

Fig 2 .
Fig 2. MARC1 A165T substitution leads to reduction in MARC1 protein expression.(A) Plasmids with empty vector (Vector), wild-type and common variants of human MARC1 (pTK promoter, C terminal 3x Flag tag) were expressed in HuH-7 cells.48 hours after transfection, cells were split into two parts.First part of cells were solubilized in RIPA buffer, proteins separated on SDS-PAGE (4-20%) gels and blotted with corresponding antibodies, (B) immunoblot quantitation was performed.(C) Second part of cells was used to measure MARC1 mRNA levels using Real-Time PCR.(D) Human liver samples DNA were extracted, and human MARC1 gene third exon that contains MARC1 variants p.165A/T and p.187M/K was amplified by PCR and Sanger sequenced, and samples with homozygous MARC1 p.187M variant selected.Liver samples proteins were extracted, size fractioned and blotted as described in Methods.(E) Western Blot signals were measured with ECL imager build-in software, (F) RNA was extracted from the same livers, and MARC1, MARC2 and house-keeping gene (36B4) expression levels were detected by qRT-PCR.For human MARC1 primary antibody control MARC1 deficient HuH-7 cell line (CAS9/Crispr) was used as negative control.For MARC2 primary antibody control MARC1 deficient HuH-7 cell line (expressing no MARC1 and 2) was used as negative control and the same cell line with transient overexpression of recombinant human MARC2 without tags, as positive control.Red arrowsreduction (%) of MARC1 protein.Mean ± s.e.m. are shown in all graphs, **p<0.01.https://doi.org/10.1371/journal.pgen.1011179.g002 was inactivated by DNA homologous recombination, replacing Marc1 exon 2 and 3 with antibiotic selection marker, β-galactosidase gene (LacZ), Cre recombinase (self-deleting protamine promoter), as well as with 2 Cre sites.After the first expression of new construct in mouse spermatids antibiotic selection marker and Cre recombinase were deleted by Cre recombination, leaving the β-galactosidase gene (LacZ) and one LoxP site in the flanking intron (Fig 5A).The subsequently generated Marc1 -/-(Marc1 KO) mice were born live with the expected Mendelian mode of inheritance and displayed no overtly abnormal phenotype.Adult Marc1 KO were fertile and nursed their pups normally (S4 Table).Full deletion of Marc1 gene in Marc1 homozygous KO mouse was confirmed in liver on DNA (PCR), mRNA (qRT-PCR) and protein levels (Western Blot) (Fig 5B-5D).Marc2 paralogue or Marc1/2 cofactors (Cyb5B and Cyb5r3) proteins expression were not affected by Marc1 deletion (Fig 5D).

Fig 3 .
Fig 3. Human MARC1 p.165T variant causes mislocalization of MARC1 protein in hepatic cells.HuH-7 cells were transfected with plasmids containing wild type or 2 common genetic variants (p.A165T and p.M187K) of human MARC1 (with C terminal Flag tag and gene expression driven under CVM promoter).48 hours after transfection cells were incubated with Mitotracker Orange CMTMRos staining (Thermo Fisher) solution in serum free media for 20 minutes, cells were washed, fixed, permeabilized and blocked as described in Methods.After additional washing, cells were incubated with Flag antibody (Sigma) for 1 hour, washed and incubated with Alexa Fluor 647 conjugated secondary antibody.After final wash, cells were incubated in mounting medium with DAPI (Ibidi) and imaged with Leica confocal microscope.The experiment was repeated once with the equivalent results.https://doi.org/10.1371/journal.pgen.1011179.g003

Fig 4 .Fig 5 .
Fig 4. Marc1 and Marc2 have distinct mRNA and protein expression patterns in mouse tissues.Chow fed 8-10 weeks old wild type C57BL/ 6N male mice (n = 10) were sacrificed at ad lib fed state, tissues collected, and Marc1 (A) and Marc2 (B) mRNA gene expression detected by RNAseq as described in Methods.Data plotted as mean with standard error.From some tissues only 3-6 samples were analyzed.Values were calculated as Transcripts Per Million (TPM).(C) 12 weeks old wild type C57BL/6N male mice fed ad lib chow diet were sacrificed and represented tissues collected.Tissue proteins were extracted, 20 μg of total protein of each organ were size fractionated by SDS-PAGE (4-20%)

Fig 6 .
Fig 6.Marc1 gene deletion has no effect on circulating or liver lipids in male mice fed High fat high fructose diet (HFHFD).Eight weeks old male mice (10 WT and 14 Marc1 KO) were fed with HFHFD for 34 weeks.(A) Body mass gain was measured every 2 weeks, (B) body composition (Micro CT) in mice were assessed at week 13 of the experiment.(C) Serum, (D) liver lipids, (E) selected liver gene expression (mRNA by qRT-PCR) were measured at the end of experiment.Each point represents individual mouse, mean ± s.e.m. are shown in all graphs.****p<0.0001.https://doi.org/10.1371/journal.pgen.1011179.g006

Fig 7 .Fig 8 .
Fig 7. Marc1 gene deletion has no effect on liver lipid accumulation, inflammation or fibrosis in mice fed High fat high fructose diet (HFHFD).(A) Liver samples were collected from mice described in Fig 6. at the end of the experiment.Fixed liver sections were stained with hematoxylin/eosin (HE), CD45 and Sirius Red staining.(B) Lipid droplet area, (C) CD45 positive cell area and (D) fibrosis area were measured from each animal liver section.Slides were scanned on Aperio AT2 scanners (Leica Biosystems) with 10X magnification.Each point represents individual mouse, mean ± s.e.m. are shown in all graphs.https://doi.org/10.1371/journal.pgen.1011179.g007