Hyperphosphorylation of hepatic proteome characterizes nonalcoholic fatty liver disease in S-adenosylmethionine deficiency

Summary

Methionine adenosyltransferase 1a (MAT1A) is responsible for hepatic S-adenosyl-L-methionine (SAMe) biosynthesis. Mat1a^−/− mice have hepatic SAMe depletion, develop nonalcoholic steatohepatitis (NASH) which is reversed with SAMe administration. We examined temporal alterations in the proteome/phosphoproteome in pre-disease and NASH Mat1a^−/− mice, effects of SAMe administration, and compared to human nonalcoholic fatty liver disease (NAFLD). Mitochondrial and peroxisomal lipid metabolism proteins were altered in pre-disease mice and persisted in NASH Mat1a^−/− mice, which exhibited more progressive alterations in cytoplasmic ribosomes, ER, and nuclear proteins. A common mechanism found in both pre-disease and NASH livers was a hyperphosphorylation signature consistent with casein kinase 2α (CK2α) and AKT1 activation, which was normalized by SAMe administration. This was mimicked in human NAFLD with a metabolomic signature (M-subtype) resembling Mat1a^−/− mice. In conclusion, we have identified a common proteome/phosphoproteome signature between Mat1a^−/− mice and human NAFLD M-subtype that may have pathophysiological and therapeutic implications.

Graphical abstract

Highlights

• •SAMe deficiency in mouse and human NAFLD causes hyperphosphorylation in the liver
• •Hyperphosphorylation correlates with activation of the kinases, CK2α and AKT1
• •SAMe deficiency alters mitochondrial, peroxisomal, and protein translation pathways
• •SAMe-deficient phosphoproteome signature may be therapeutically relevant in NAFLD

Human metabolism; Molecular biology; Proteomics.

Introduction

Nonalcoholic fatty liver disease (NAFLD) has emerged as the most common cause of chronic liver disease with worldwide prevalence of 25% that reflects the global rise in obesity.¹ NAFLD encompasses a spectrum, from simple steatosis to nonalcoholic steatohepatitis (NASH), a condition characterized by the concordance of steatosis with liver injury, inflammation, and fibrosis.^1,2,3 Roughly 25% of patients with NAFLD have NASH,³ of whom 25% can progress to liver cirrhosis, of whom 1%–2% will develop hepatocellular carcinoma in their lifetime.³ There are currently no FDA-approved treatments for NASH, and consequently, NASH has recently overtaken hepatitis C as the most common indication for liver transplantation for woman in the United States, highlighting the need for a viable treatment for this major public health problem.^4,5

To understand disease progression of NASH in humans as well as to identify molecular targets involved in the disease progression, murine models are commonly used as they mimic some but not all aspects of human disease.^6,7 Recent work has shown that nearly half of patients with NAFLD have a serum metabolomic profile (M-subtype) that closely resembles the methionine adenosyltransferase 1a knockout mouse (Mat1a^−/−) model.⁸ Prior to any histological signs of disease, four-month-old Mat1a^−/− mice (pre-disease Mat1a^−/−) are more susceptible to diet-induced liver steatosis.⁹ By eight months of age, Mat1a^−/− mice spontaneously develop NASH (Mat1a^−/− NASH), which can be reversed by exogenous S-adenosyl-L-methionine (SAMe) administration histologically and biochemically.^8,9 SAMe is the principal methyl donor responsible for DNA, RNA, and protein methylation.¹⁰ In the liver, majority of the biosynthesis of SAMe is via MAT1A-encoded isoenzyme while MAT2A synthesizes SAMe in non-hepatic tissues.¹¹ In the Mat1a^−/− NASH model, kinase/phosphatase signaling is dysregulated but the impact on protein phosphorylation (and the proteome) is unknown. Mat1a^−/− animals have higher activity of ERK2,¹² LKB1,^13,14 AKT,¹⁴ and AMPK^13,14 in addition to decreased dual-specificity MAPK phosphatase 1 (DUSP1) activity¹² suggestive of increased phosphorylation with disease. Importantly, exogenous SAMe administration normalized ERK2 and DUSP1 activity in the Mat1a^−/− NASH model,¹² suggesting that SAMe administration should normalize hyperactive kinases in a SAMe-deficient system.

The goals of this study were: (1) quantify the proteome (protein quantity) and phosphoproteome (site-specific phosphorylation quantity) signatures at different stages in the progression of the Mat1a^−/− NASH mouse model to gain insight into the temporal order of altered signaling pathways, (2) to identify SAMe-sensitive signatures linked to the beneficial outcome of SAMe treatment, and (3) to identify conserved disease signatures in human M-subtype human NAFLD that are also present in the Mat1a^−/− model. Furthermore, early alterations during disease progression can provide clues on the initiating pathogenesis of disease and provide early biomarkers of disease. Our analyses revealed a global hyperphosphorylation that is already striking in the pre-disease Mat1a^−/− livers that widely normalized after SAMe administration. The phosphoproteome revealed a hyperactive casein kinase 2α (CK2α) and AKT1 signature in the Mat1a^−/− NASH model that is responsive to SAMe administration. Importantly, the same hyperphosphorylation and hyperactive CK2α and AKT1 signature were observed in human NAFLD with the M-subtype.

Results

To understand the molecular underpinnings of and to identify early drivers in the Mat1a^−/− NASH model, we utilized two groups of animals at, one “pre-disease” and the other with NASH (referred to as Mat1a^−/− NASH). The first cohort of mice consists of Mat1a^−/− mice and wild-type (WT) littermates aged four months. We showed that the livers from mice around this age are normal, but they developed massive fatty liver when challenged with a choline-free diet for only six days.⁹ These mice were given 100 mg/kg/day of SAMe for seven days (Figure S1A), which we had shown normalized hepatic SAMe level in Mat1a^−/− mice.¹² Histological evaluation, triglyceride levels, and alanine transaminase (ALT)/aspartate transaminase (AST) levels for this cohort are shown in Figures S1B–S1D, respectively. We also studied a cohort of Mat1a^−/− animals aged ten months with NASH. This cohort includes mice given 30 mg/kg of SAMe five days a week for eight weeks starting at 8 months of age (Figure S1E). We demonstrated that SAMe administration in this cohort nearly normalized the liver histology and ALT/AST levels.⁸

To identify potential kinase and phosphatase signaling pathways and the downstream phosphorylated proteins, we performed a TiO2 phospho(tryptic) peptide affinity enrichment of both the pre-disease and NASH cohorts of Mat1a^−/− animals and performed data-dependent acquisition-mass spectrometry analysis. In addition, total protein quantity was obtained through data-independent acquisition-mass spectrometry from the same samples. The results from these experiments and % fold change normalized to total protein (phosphorylation/total protein) are shown in Table 1. There was a dramatic increase in phosphorylation sites in Mat1a^−/− pre-disease and NASH livers compared to age- and gender-matched WT livers with 525 and 685 phosphosites altered, which corresponded to an increase in 364 and 528 phosphoproteins, respectively (Figures 1A and 1B, Table 1). Normalizing phosphoproteome to the specific protein quantity allowed assessment of whether the changes in phosphorylation status are due to i) altered phosphorylation at specific sites, or ii) a change in the protein quantity. As shown in Table 1, regardless of disease state, majority of the changes were due to an increase in the phosphorylation at each phosphosite (90.4% and 87.9% hyperphosphorylation for Mat1a^−/− pre-disease and NASH, respectively). This pathological signature corresponds to 52 hyperphosphorylated proteins in pre-disease Mat1a^−/− and 148 hyperphosphorylated proteins in Mat1a^−/− NASH. Raw data files of Mat1a^−/− pre-disease and NASH are uploaded on (https://panoramaweb.org/Mat1a_NASH.url, ProteomeXchange ID: PXD022122, https://panoramaweb.org/Larp1.url).

Table 1.

Summary of phosphoproteomics

Sample Information		Phosphorylation Raw Data				Phosphorylation/Total Protein Data
Group Comparison	Disease Stage	Total Phospho-sites Increased	Total Phospho-sites Decreased	Phospho-sites Increased (significant)	Phospho-sites Decreased (significant)	Total Phospho-sites Increased	Total Phospho-sites Decreased	Phospho-sites Increased (significant)	Phospho-sites Decreased (significant)
Mat1a−/− vs. WT	Pre-disease	2838	726	525	35	902	271	66	7
Mat1a−/− + SAMe vs. Mat1a−/−	Pre-disease	1426	2210	60	112	317	909	13	116
Mat1a−/− vs. WT	NASH	2809	1500	685	232	1500	582	269	37
Mat1a−/− + SAMe vs. Mat1a−/−	NASH	827	3395	117	1081	434	1526	28	303
Human M-Type vs. Non-M	NAFLD	753	161	232	20	648	470	69	41

The number of phosphosites before and after normalization to total protein levels was compared between groups of Mat1a−/− pre-disease/NASH and human M/non-M-type NASH. Significance is defined as the phosphosites in each group with p < 0.05.

Figure 1.

Phospho-proteomic changes in Mat1a^−/− mice livers

(A) Phospho-proteomic changes found in pre-disease Mat1a^−/− animals compared to WT littermates aged four months (n = 4/condition). 252 phosphopeptides in this comparison with a p value <0.05 and a fold change greater than 25% in either direction. Of these phosphopeptides, 236 are increased in Mat1a^−/− while only 16 phosphopeptides are decreased. The X axis denotes the gene name for each protein and their corresponding phosphosite.

(B) Phosphoproteomic changes found in Mat1a^−/− NASH compared to WT littermates aged ten months (n = 6/condition). 917 phosphopeptides in this comparison with a p value <0.05 and a fold change greater than 25% in either direction. Of these phosphopeptides, 685 are increased in Mat1a^−/− while 232 phosphopeptides are decreased in Mat1a^−/−.

(C) IPA phosphorylation analysis of predicted activation status of upstream kinases through a curated database of phosphoproteomic datasets found CK2α was predicted to be activated in pre-disease Mat1a^−/− (activation Z score = 4.49, adjusted p value = 5.88E-7) and in Mat1a^−/− NASH (activation Z score = 3.394, adjusted p value = 2.34E-14). The left node of the bicolor node represents the pre-disease condition and the right node represents the NASH condition. The color gradient of the bicolor nodes is the range of log2-fold changes from predicted inhibition (blue gradient) to activation (orange gradient).

(D) IPA phosphorylation analysis found AKT1 was predicted to be activated in pre-disease Mat1a^−/− (activation Z score = 2.83, adjusted p value = 1.06E-3) and in Mat1a^−/− NASH (activation Z score = 1.149, adjusted p value = 2.61E-2). The bicolor nodes are defined as in “C” above.

Sixty of the differentially expressed hyperphosphorylated proteins in the Mat1a^−/− NASH livers are established downstream phosphotargets of both CK2α and AKT1. The results are consistent with the activation of both kinases at all disease stages in the Mat1a^−/− NASH model (Figures 1C and 1D), In pre-disease stage, although there was a non-significant increase in CK2α protein levels (18.3%, 11.3% false discovery rate or FDR), CK2α and AKT1 activation was predicted based on phosphoproteomics results (Figures 1C and 1D), and we quantified an increase in AKT1 Ser129 phosphorylation (157% increase, p value = 0.0073). In Mat1a^−/− NASH compared to WT animals, there was an increase in CK2α protein expression (57%, 0.00001% FDR). Increased CK2α and AKT1 activity was predicted from phosphoproteomics (Figures 1C and 1D), although little or no AKT1 Ser129 phosphorylation in Mat1a^−/− NASH was observed (39.1% increase, 0.19 p value). A well-known target of CK2α is α-Catenin, which is phosphorylated by CK2α at Ser641 residue regulating its interaction with β-catenin.¹⁵ Both the pre-disease and NASH Mat1a^−/− phosphoproteome showed a 1.4- to 1.7-fold induction in Ser641 α-catenin phosphorylation (both are significantly different from WT with p < 0.005) (Figure 1C).

Phosphorylation changes can be regulated by phosphatases, and within the protein quantity and phosphorylation datasets, there are several dysregulated serine, threonine, and dual specificity phosphatases which we focused on as the majority of phosphorylation changes occurred on these amino acid residues (Tables S1 and S2). There was minimal overlap in the altered phosphatases in pre-disease and Mat1a^−/− NASH with only one (PPM1F) in common. Overall, there were more phosphatases inhibited than activated (4/6 and 6/8, respectively) including decreases in PPA2, MDP1, SHOC2, and PPM1F in pre-disease Mat1a^−/− livers compared to respective WT controls (Table S1). In Mat1a^−/− NASH, six phosphatases were decreased compared to WT (Table S2) including CPPED, a phosphatase known to deactivate AKT by dephosphorylating AKT Ser473¹⁶ and protein phosphatase 5C (PPP5C), a known inhibitor of ASK1 through its phosphatase activity.¹⁷

Although the hyperphosphorylation signature was dominant in Mat1a^−/− animals, there were also quantitative changes in proteins involved in lipid metabolism. Of the 411 differentially expressed proteins (Figure 2A) in pre-disease Mat1a^−/−, there were 24 peroxisomal proteins (adjusted p value = 2.83E-23), 6 are involved in peroxisomal lipid metabolism (adjusted p value = 6.95E-06), 25 lysosomal proteins (adjusted p value = 5.70E-13), and 141 mitochondrial proteins (adjusted p value = 8.15E-107), 10 are involved in mitochondrial fatty acid β-oxidation (adjusted p value = 2.09E-11) (Figures 2B and S2, Table S3). In Mat1a^−/− animals with NASH, there is twice the number of differential proteins compared to pre-disease (921 verse 411, respectively, Figure 2C). Interestingly, there is a similar set of cellular compartments and pathways altered, specifically lipid metabolism (17.6% enrichment), 310 mitochondria proteins (adjusted p value = 4.12E-249), 12 proteins involved in mitochondrial fatty acid β-oxidation (adjusted p value = 6.74E-11) and 35 peroxisomal proteins (adjusted p value = 2.80E-29) with 11 involved in peroxisomal lipid metabolism (adjusted p value = 3.35E-11). In addition, in Mat1a^−/− NASH, there was also an additional 339 nuclear proteins (adjusted p value = 3.23E-102) and 14 spliceosomal proteins (adjusted p value = 0.00005), both cellular compartments important for transcription (38 proteins, adjusted p value = 5.46E-9) and translation (98 proteins, adjusted p value = 3.20E-69) (Figures 2D and S3, Table S3). Taken together, early in disease before visible steatosis, there is already a reduction in the expression of proteins involved in mitochondrial β-oxidation and an increase in peroxisomal β-oxidation. This pattern continues with the onset of NASH. In addition, NASH also exhibits alterations in the transcriptional and translational proteome. For transcription or translation, a definitive direction could not be deciphered. However, a specific subset of ribosomal proteins and translation factors found to be responsive to the translational regulator; LARP1 were found to be induced in Mat1a^−/− NASH mice as we reported previously.¹⁸ Consistently, immunofluorescent staining in Mat1a^−/− NASH livers shows increased peroxisome size without a change in number, which was normalized by SAMe treatment (Figure S6B). However, mitochondrial size and number were unchanged in the Mat1a^−/− NASH livers (Figure S6C).

Figure 2.

Proteomic changes in Mat1a^−/− livers

(A) 411 proteins (FDR <1% and a fold change greater than 25% in either direction) were found in pre-disease Mat1a^−/− animals compared to WT littermates aged four months (n = 4/condition). Of these proteins, 166 are increased in Mat1a^−/− while 245 are decreased.

(B) ClueGO Ontology Analysis via PINE¹⁹ for visualization of altered KEGG Cellular Components in pre-disease Mat1a^−/− animals compared to WT littermates aged four months. Size of node denotes p value of enrichment.

(C) 921 proteins were found in Mat1a^−/− animals aged ten months with NASH compared to WT littermates (n = 6/condition). Of these proteins, 294 are increased in Mat1a^−/− while 627 are decreased.

(D) ClueGO Ontology Analysis via PINE for visualization of altered KEGG Cellular Components in Mat1a^−/− animals aged ten months with NASH compared to WT littermates. Size of node denotes p value of enrichment.

In pre-disease Mat1a^−/− animals, although we could not detect enrichment in altered nuclear protein or proteins involved in translation, we observe changes in the phosphoproteome of these animals, which foreshadow these altered cellular compartments and pathways seen in Mat1a^−/− NASH. In pre-disease Mat1a^−/− animals, there are altered phosphoproteins which reside in the ribonuclear protein granule (15 proteins, adjusted p value = 9.79E-10) and involved in translation (31 proteins, adjusted p value = 1.69E-17) (Table S3) and the peroxisome (five proteins, adjusted p value = 0.012), including four peroxins which are involved in peroxisomal protein import and peroxisomal biogenesis.²⁰ Equally, phosphoproteins involved in lipid metabolism (ACACA, ACACB, PRKAB2, and PRKAG2, adjusted p value = 0.00027) were altered in pre-disease conditions. Alterations in phosphoproteins involved in regulation of fatty acid oxidation/synthesis (ACACB, PRKAB2, and PRKAG2, adjusted p value = 0.037), lipid homeostasis (9 proteins, adjusted p value = 0.0012), and lipid metabolism (15 proteins, p value = 3.36E-11) were observed in Mat1a^−/− NASH. Taken together, Mat1a deficiency had an overall effect of lowering fatty acid oxidation and enhancing fatty acid synthesis (Table S3).

To understand the effect of how low cellular methylation capacity in Mat1a^−/− changes the proteome, we next examined the proteomic changes after SAMe administration at different disease stages. Administration of SAMe to the pre-disease animals resulted in substantial (65%) reversal of phosphorylation with (112/172) phosphopeptides decreased with SAMe treatment (Figure 3A, Table 1). In Mat1a^−/− NASH, there was an even stronger reversal (90%) in hyperphosphorylation signature of phosphorylated peptides (1081 of the 1109 sites) decreased with SAMe treatment (Figure 3D and Table 1). These phosphorylation changes were normalized to the total protein quantity in pre-disease Mat1a^−/− and Mat1a^−/− NASH (Table 1). The reversal in phosphorylation by SAMe treatment is widespread which suggests that there is global deactivation of kinases and/or activation of phosphatases. Specifically, enrichment in established downstream phosphotargets of AKT1 shows activation in pre-disease Mat1a^−/− livers that were normalized by treatment with SAMe (Figure 3C). There is also inhibition in CK2α phosphotargets, such as residue Ser641 of α-catenin in SAMe-treated pre-disease Mat1a^−/− compared to pre-disease Mat1a^−/− alone, consistent with an inhibition and normalization of CK2α (Figures 3B and S4). Furthermore, SAMe administration decreased (25%, p value = 0.012) phosphorylation of AKT1 at Ser¹²⁹, a known activating residue phosphorylated by CK2.^21,22,23 In Mat1a^−/− NASH, ingenuity pathway analysis also predicted that SAMe treatment inhibits both AKT1 and CK2α and normalizes their downstream phosphotargets. Consistently, SAMe administration inhibited both CK2α protein levels (32% compared to WT control, 7.95E-8 FDR) (Figure 3E) and AKT1 Ser129 (51.7% compared to WT control, p value = 0.04) (Figure 3F). In pre-disease Mat1a^−/− mice, SAMe treatment enhanced the levels of the phosphatase, CPPED, by 44% compared to Mat1a^−/− KO (0.027 FDR) and normalized CPPED to WT levels. In NASH condition, SAMe treatment caused an 11.1% induction in CPPED compared to Mat1a^−/− KO (0.289 FDR) and normalized CPPED to WT levels. In Mat1a^−/− NASH, SAMe administration also raised PPP5C level by 25% (FDR = 0.12) (Table S4). Additionally, based exclusively on our phosphoproteomics analysis, activation was predicted for PPP1CA (activation Z score = −2.425, adjusted p value = 5.99E-3) and PTPN11 (activation Z score = −2.456, adjusted p value = 2.26E-2) with SAMe administration in Mat1a^−/− NASH (Figure S5).

Figure 3.

Phospho-proteomic changes with SAMe treatment in Mat1a^−/− livers

(A) 172 phosphopeptides (FDR <1% and a fold change greater than 25% in either direction) were found in in pre-disease Mat1a^−/− animals treated with 100 mg/kg of SAMe for seven days were compared to Mat1a^−/− animals treated with PBS for seven days (n = 4/condition). Of these phosphopeptides, 60 are increased with SAMe treatment while 120 are decreased.

(B) Phosphorylation analysis shows downstream CK2α targets which are normalized with SAMe treatment in pre-disease Mat1a^−/−. The left node of the bicolor node represents the pre-disease condition, and the right node represents the pre-disease condition treated with SAMe. The color gradient of the bicolor nodes is the range of log2-fold changes from predicted inhibition (blue gradient) to activation (orange gradient).

(C) IPA phosphorylation analysis found AKT1 was predicted to be decreased with SAMe treatment and normalized in pre-disease Mat1a^−/− (activation Z score = −1.57, adjusted p value = 5.36E-3). The bicolor nodes are defined as in “B” above.

(D) 1198 phosphopeptides (FDR <1% and a fold change greater than 25% in either direction) were found in in Mat1a^−/− animals with NASH treated with 30 mg/kg of SAMe for five days a week for eight weeks as compared to Mat1a^−/− animals treated with PBS (n = 5/condition). Of these proteins, 117 are increased with SAMe treatment in Mat1a^−/− while 1081 are decreased.

(E) IPA phosphorylation analysis found CK2α was predicted to be inhibited and normalized with SAMe treatment in Mat1a^−/− NASH (activation Z score = −4.715, adjusted p value = 6.89E-12). ^−/−. The left node of the bicolor node represents the NASH condition, and the right node represents the NASH condition treated with SAMe. The color gradient of the bicolor nodes is the range of log2-fold changes from predicted inhibition (blue gradient) to activation (orange gradient).

(F) IPA phosphorylation analysis found AKT1 was predicted to be inhibited and normalized with SAMe treatment in Mat1a^−/− NASH (activation Z score = −3.813, adjusted p value = 1.43E-2). The bicolor nodes are defined as in “E” above.

To better understand the impact of SAMe administration on the total proteome (Table 1) in the pre-disease (134 proteins, Figure 4A) and in the Mat1a^−/− NASH mode (379 proteins, Figure 4C), we focused on the SAMe-sensitive differentially expressed proteins. In pre-disease animals, peroxisomal proteins dominated, with 26/29 (89%) decreased with SAMe administration (adjusted p value 4.25E-47, Figure 4B). This also occurred in Mat1a^−/− NASH but in addition to peroxisomal proteins there were SAMe-sensitive enrichment in proteins belonging to the mitochondria, ribosome, and ER subproteomes (Figures 4D and S6) and a decrease in mitochondrial β-oxidation components that were normalized by SAMe treatment (Figure S6).

Figure 4.

Proteomic changes with SAMe administration in Mat1a^−/− livers

(A) 134 proteins (FDR <1% and a fold change greater than 25% in either direction) were found in pre-disease Mat1a^−/− animals treated with 100 mg/kg of SAMe for seven days as compared to Mat1a^−/− animals treated with PBS for seven days (n = 4/condition). Of these proteins, 71 are increased with SAMe administration while 63 are decreased with the addition of SAMe.

(B) Upstream regulator analysis using IPA found a strong predicted activation of PPARA with SAMe administration to pre-disease Mat1a^−/− (Z score = 3.379, p value = 1.08E-27) which normalizes PPARA.

(C) Cystoscope visualization of proteins from the most enriched KEGG Cellular Components in pre-disease Mat1a^−/− treated with SAMe vs Mat1a^−/− generated using PINE.¹⁹ Fold changes of proteins are represented as colors from blue to red.

(D) ClueGO Ontology Analysis via PINE for visualization of altered KEGG Cellular Components in Mat1a^−/− animals aged ten months with NASH treated with SAMe compared to animals treated with PBS. Size of node denotes p value of enrichment.

Next, we determined whether there was conservation of the disease signature between Mat1a^−/− NASH model and human NAFLD and found synergy with widespread hyperphosphorylation of the proteome and protein abundance changes in the mitochondria and the peroxisome subproteomes. Based on phosphoproteomics analysis of 13 liver biopsies from patients with NAFLD with various levels of steatosis and fibrosis (see Table 2 for clinical characterization), there were two major clusters and although they did not correlate with the degree of steatosis or fibrosis (Figures 5A and 5B), one cluster was similar to Mat1a^−/− mice with respect to the phosphoprotein signature. Thus, we denoted the group that is similar to Mat1a^−/− mice M-subtype (n = 10) and the other group non-M subtype (n = 3). Seven of the 10 M-subtype patients were confirmed to have serum metabolomic and lipidomic signature that resemble Mat1a^−/− NASH.⁸ As found with the Mat1a^−/− NASH model, there was a global hyperphosphorylation signature (normalized to total protein levels) in the M-subtype versus non-M human NAFLD, with 71% (649/914) phosphopeptides higher in the M-subtype group. Intriguingly, CK2α was predicted to be activated in the M-subtype (activation Z score 4.382, adjusted p value 1.08E-14) and the CK2α target, phospho-S641-α-catenin, exhibited a 2-fold induction (FDR = 0.167) (Figure 5C), similar to the observed CK2α activation in both pre-disease and Mat1a^−/− NASH (Figure 5D). These samples also show predicted activation of AKT1, like pre-disease Mat1a^−/− animals (Figure 5D). Furthermore, there was an enrichment in differentially expressed mitochondrial, ribosomal, spliceosomal, and proteasomal proteins in M-subtype NAFLD when compared to non-M NAFLD (Figure 6B). These altered organelles have significant overlap with Mat1a^−/− NASH and suggest a similar pathophysiology in Mat1a^−/− NASH and human M-subtype NAFLD.

Table 2.

Subtyping human NASH

Sample Information		NASH Stages				Clinical Characteristics				Subtyping Results
Sample ID	Collection Date	F0	F1	F2	F3	Steatosis	Lobular inflammation	Hepatocyte Ballooning	Fibrosis	Metabolomics Subtype	Phospho-Proteomics Subtype
NAFLD9	5/2/2019			X		∼25%	not identified	not identified	2	Type M	Type M
NAFLD11	12/12/2017		X			25%	not identified	not identified	1a	Type M	Type M
NAFLD13	3/26/2018		X			10%	scant	not identified	1a	Type M	Type M
NAFLD2	3/26/2018		X			30%	not identified	not identified	1a	Type M	Type M
NAFLD10	5/3/2018		X			10%–15%	mild, lymphoid, focal	not identified	1a	Type M	Type M
NAFLD3	12/29/2017			X		<10%	mild	not identified	2	Type M	Type M
NAFLD7	5/22/2018				X	40%	scant	not identified	3	Type M	Type M
NAFLD1	10/13/2017		X			20%–30%	not identified	not identified	1a		Type M
NAFLD8	2/27/2018		X			40%	not identified	not identified	1a		Type M
NAFLD12	10/31/2019		X			10%–20%	Few ceroid macrophages	not identified	1a		Type M
NAFLD6	8/28/2018		X			30%	not identified	rare	1a		Non-M
NAFLD4	7/26/2019		X			60%	not identified	not identified	1a		Non-M
NAFLD5	2/2/2018		X			50%	patchy glycogenic nuclei	not identified	1a		Non-M

Figure 5.

Phospho-proteomic changes in human NAFLD with comparisons to Mat1a^−/− KO model

(A) Principal component analysis of phosphoproteomic changes acquisitions from human NASH. Original values are ln(x)-transformed. Unit variance scaling is applied to rows; SVD with imputation is used to calculate principal components.

(B) Heatmap of human phosphoproteomics data. Rows were centered and phosphopeptide intensities were log transformed; unit variance scaling is applied to rows. Both rows and columns are clustered using Manhattan distance and average linkage.

(C) IPA phosphorylation analysis found CK2α predicted to be activated in the M-subtype NASH when compared to non-M subtype NASH (activation Z score 4.382, adjusted p value 1.08E-14) (activation Z score = 2.83, adjusted p value = 1.06E-3).

(D) IPA phosphoproteomic upstream regulator analysis comparing activation status of predicted upstream regulators in pre-disease Mat1a^−/− animals, Mat1a^−/− animals with NASH, and human M-subtype NASH.

Figure 6.

Proteomic changes in human NAFLD

(A) 974 proteins (FDR <1% and a fold change greater than 25% in either direction) were found in M-subtype NAFLD when compared to non-M subtype NAFLD (M-subtype NAFLD, n = 10, Non-M subtype, n = 3). Of these proteins, 424 are increased in M-subtype NAFLD while 550 are decreased in M-subtype NAFLD.

(B) ClueGO Ontology Analysis via PINE¹⁹ for visualization of altered KEGG Cellular Components in human M-subtype NAFLD vs non-M subtype NAFLD. Size of node denotes p value of enrichment.

Similar to Mat1a^−/− livers, there was a subset of dysregulated serine and threonine and dual-specificity phosphatases that show decreased expression in M-subtype compared to non-M subtype NAFLD (8/12). These include DUSP23, PP5C, and PTPA that were unique to human and PPM1F that was similar to the animal model (Table S5). Taken together, M-subtype human NAFLD also have lower expression of multiple phosphatases and they may contribute to the observed global hyperphosphorylation.

Discussion

Analyzing the proteome and phosphoproteome at different stages of progression of the Mat1a^−/− NASH model has provided insight into the temporal order of the cellular pathways associated with disease progression. Early pre-disease animals primarily exhibited alterations in the mitochondria, peroxisome, and lysosome subproteomes with specific induction in peroxisomal β-oxidation and a decrease in mitochondrial β-oxidation components. This may represent the stress on lipid metabolism machinery to metabolize excess lipids before visible steatosis (Table S3). Mitochondrial dysfunction in NASH is a well-known phenomenon^24,25 and our group has recently reported the role of MAT1A’s influence on mitochondrial function.²⁶ In addition to the pathways altered in the pre-disease state, the NASH stage also exhibits dysregulation of ER proteins, cytosolic ribosomes, spliceosomes, and nuclear proteins that are associated with the onset of steatosis and subsequent inflammation and fibrosis and transition to hepatocellular carcinoma^27,28,29 (Table S3).

There is a global hyperphosphorylation signature at all stages of the Mat1a^−/− NASH model which is consistent with CK2α activation. CK2α is the catalytic subunit of casein kinase 2 (CK2) that is a ubiquitously expressed, constitutively active.^30,31 CK2 is known to be involved in cell cycle progression³⁰ and is overexpressed in HCC.³² CK2 regulates a multitude of pathways including AKT1,^21,33 NF-κB,³⁴ TGF-β,³⁵ ERK2, β-catenin, and WNT signaling.¹⁵ CK2 activation has been reported in type 2 diabetes³⁶ and obesity where it was shown to regulate lipogenesis and triglyceride levels³⁷ and AKT1-mediated glucose metabolism and insulin signaling by direct phosphorylation of AKT1 Ser129 and indirectly through PTEN inhibition.^22,23,33, CK2 was recently shown to be increased in patients with NAFLD, with its expression correlating to disease severity.³⁸ Importantly, we found in pre-disease and Mat1a^−/− NASH CK2α and AKT1 activation. Previous reports have shown that AKT1 activation in a SAMe-deficient system is due to LKB1 and that Mat1a^−/− animals have increased LKB1 activity.¹⁴ We confirmed increased LKB1 (STK11 Ser31) phosphorylation in pre-disease Mat1a^−/− animals but our current findings also implicate direct phosphorylation of AKT1 Ser129 by CK2 in the Mat1a^−/− NASH model at all disease stages.

Through global phosphoproteomics, insights were gained on the broad effects of SAMe administration on the proteome in the Mat1a^−/− NASH model. We observed a drastic reduction and normalization of the phosphoproteome with 90% of the differentially expressed phosphopeptides decreased with SAMe administration in Mat1a^−/− NASH (Figure 3C). This phosphorylation signature is representative, at least in part, of CK2α and AKT1 normalization (Figure 3). Mat1a^−/− mice are known to have higher β-catenin activity and SAMe administration was shown to inhibit β-catenin activation in liver and colon cancer cells³⁹ but the underlying mechanism was unclear. SAMe administration also normalized hyperactive ERK2 in the Mat1a^−/− livers¹² and reversed disease progression in Mat1a^−/− NASH.⁸ CK2 regulates β-catenin through WNT-dependent and independent signaling. One mode of control of β-catenin is via its interaction with α-catenin that is known to inhibit β-catenin’s transcriptional activity in cancer cells.¹⁵ Growth factor-mediated ERK activation enhances CK2 activity that in turn phosphorylates α-catenin at Ser641 residue,¹⁵ which inhibits α-catenin’s interaction with β-catenin leading to activation of β-catenin transcriptional activity. SAMe administration reversed hyperphosphorylation of Ser641-α-catenin in both Mat1a^−/− pre-disease and NASH, suggesting this may be a key mechanism for SAMe to inhibit β-catenin signaling in this model. Importantly, we also found Ser641-α-catenin hyperphosphorylation in human M-subtype NAFLD versus non-M-subtype NAFLD. Augmented β-catenin-dependent transcription is also dependent on phosphorylation of AKT at Ser129 by CK2.²² CK2α-enhancing effects on β-catenin transcriptional activity are reversed when an AKT mutant deficient in Ser129 phosphorylation by CK2 is co-expressed.²² Our data showing AKT1 Ser¹²⁹ phosphorylation induced in Mat1a^−/− pre-disease and NASH and normalized by SAMe support the activation of CK2-β-catenin pathway can be normalized by SAMe treatment (Figures 1C and 1D).

Regulation of CK2 signaling is not fully elucidated, as it is constitutively active³¹ and not regulated like a traditional kinase through phosphorylation of its activation loop.^40,41 In fact, recent studies have shown that modulation of CK2 activity is dependent on post-translational modifications (PTMs), and CK2-mediated phosphorylation requires precise positioning of multiple adjacent phosphorylated residues or other PTM-modified sites including methylation, acetylation, or sumoylation.^42,43 Our previous work has shown that the Mat1a^−/− NASH model has decreased acetyl-CoA and accompanying protein acetylation, as well as decreased SAMe and subsequent DNA and protein methylation.^8,44 In addition, SUMOylation is controlled by SAMe levels and that the Mat1a^−/− NASH model has increased SUMOylated RanGAP1 which is normalized by SAMe administration.⁴⁵ The proposition that the combinatorial nature of dysregulated PTMs in the Mat1a^−/− NASH might result in CK2 activation is intriguing but more work needs to be done to determine the role of each dysregulated PTM.

Global hyperphosphorylation does not occur often in biology, as phosphorylation is a tightly controlled mechanism for cellular signal transduction and is regulated by the availability of ATP, kinase activity, and levels of protein phosphatases^46,47 and suggests a coordinated multiprong signaling. Mat1a^−/− animals have higher activity of multiple kinases, such as ERK, LKB1, AKT, and AMPK^12,13,14 and lower activity of the phosphatase DUSP1.¹² In our study, increased phosphorylation in pre-disease Mat1a^−/− may partially be attributed to decreased protein phosphatases (Table S1) such as SHOC2, a regulatory subunit of PP1C, which has recently been implicated to regulate RAF-ERK via its phosphatase activity.⁴⁸

Mat1a^−/− NASH livers also have decreased protein phosphatases including serine/threonine-protein phosphatase CPPED1, which is known to dephosphorylate AKT1 Ser473.¹⁶ This could potentially contribute to increased AKT activity along with the CK2α-mediated activation of AKT through Ser¹²⁹ phosphorylation. It is challenging to pinpoint the root cause of global hyperphosphorylation in the Mat1a^−/− NASH model, as along with decreased protein phosphatases (Tables S1 and S2), global SAMe and acetyl-CoA levels are decreased causing subsequent decreases in global protein methylation and acetylation plus increased protein SUMOylation.^8,44,45 SAMe administration in the Mat1a^−/− NASH model normalizes protein phosphatases (Table S4) as well as SAMe levels and SUMOylation.^8,44 More work needs to be done to understand the temporal order of the different PTMs in the Mat1a^−/− NASH model to better understand how SAMe depletion results in these changes.

Comparison of serum metabolomics from Mat1a^−/− mice with NASH to human patients with NAFLD revealed nearly 50% of patients with NAFLD have a very similar profile as Mat1a^−/− mice, which we termed M-subtype.⁸ However, the molecular underpinnings of this subtype of human NASH have yet to be determined. By studying the phosphoproteome from liver biopsies of patients with M-subtype NAFLD, we have for the first time determined that not only do the serum lipids and metabolites of these patients resemble Mat1a^−/− serum⁸ but they also share the liver hyperphosphorylation signature highlighted by CK2α activation which is independent of the degree of steatosis and fibrosis (Figure 5C and Table 1). Taken together with CK2-mediated activation of AKT1 in the Mat1a^−/− NASH model at all disease stages, we now have a better understanding of the etiology of a major subtype of human NAFLD and further the relevancy of the Mat1a^−/− NASH model to accurately recapitulate a major subtype of human disease.

Like the Mat1a^−/− NASH model, M-subtype human NAFLD livers have decreased phosphatase activity when compared to non-M subtype NAFLD (Table S5). The decrease in PPP5C is particularly interesting because PPP5C is known to regulate ERK2,⁴⁹ a kinase we have reported to be activated in Mat1a^−/− NASH and to be SAMe responsive.¹² In addition, PPP5C has also been reported to act as a physiological inhibitor of ASK1 through its phosphatase activity.¹⁷ Strategies to inhibit ASK1 to treat NASH have shown great pre-clinical promise.^50,51 However, recent phase 3 clinical trials using selonsertib, an ASK1 inhibitor, were not able to suppress fibrosis.^52,53 Inclusion criteria for these trials did not include serum lipidomic and metabolic subtyping. Based on the results from this study and the ability to subtype human NAFLD via serum,⁸ we speculate that selonsertib may have greater success on M-subtype than non-M subtype NAFLD.

The results from this study also implore further research to be carried out using SAMe as a treatment in human M-subtype NAFLD. To date, SAMe administration in human liver disease has had mixed results but has shown benefit in intrahepatic cholestasis and a near-significant benefit in alcoholic liver disease.⁵⁴ With the ability to accurately subtype human disease via non-invasive serum metabolomic and lipidomic measurements,⁸ the use of SAMe in the patients with M-subtype NAFLD warrants further study.

In conclusion, this study has uncovered in the Mat1a^−/− model disease phophoproteome signatures occurring prior to any observable pathological phenotype and that this hyperphosphorylation signature is expanded with Mat1a^−/− NASH that was dramatically reversed with SAMe treatment. This hyperphosphorylation signature is also observed in humans with M-subtype NAFLD, which may have pathophysiological and therapeutic implications.

Limitations of the study

Our study provides a comprehensive characterization of the phosphoproteome of the liver during the development and progression of NAFLD in the setting of chronically low hepatic SAMe level. Activation of AKT and CK2 was seen in both murine model and in humans with NAFLD that share the same metabolomic signature. However, this is an association and the causal role of these kinases in the development and progression of NAFLD remains to be examined. In addition, although we identified pathways that were normalized by SAMe administration, the underlying molecular mechanisms of SAMe’s actions remain unclear.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit polyclonal anti-catalase	Abcam	Cat# ab16731; RRID:AB_302482
Rabbit monoclonal anti-VDAC1/porin + VDAC2	Abcam	Cat# ab154856; Clone [EPR10852(B)]; RRID:AB_2687466
Biological samples
Human NAFLD tissues	Biobank of the Fatty Liver Program at Cedars-Sinai Medical Center	IRB# Pro00042709
Chemicals, peptides, and recombinant proteins
S-adenosylmethionine (SAMe)	Gnosis SRL, Cairate, Italy, available via Jarrow Industries, CA	Batch code # 0-S01 CAS: 29908-03-0
Critical commercial assays
Triglyceride Colorimetric Assay Kit	Cayman Chemicals, MI	Cat# 10010303
ALT activity assay	Sigma	Cat# MAK052
AST colorimetric assay kit	Cayman Chemicals, MI	Cat# 701640
Deposited data
Panorama (https://panoramaweb.org/Mat1a_NASH.url	This paper	ProteomeXchange ID: PXD022122
Panorama (https://panoramaweb.org/Larp1.url)	Published work (Ramani et al. 202218)	ProteomeXchange ID: PXD020015
Experimental models: Organisms/strains
Mouse: Mat1a−/− KO male mice, C57BL/6 strain	Our published work (Lu et al., 20019)	N/A
Software and algorithms
MapDIA software version v.1.1	Teo et al., 201555
Skyline software-Skyline-daily (64-bit) 20.1.1.83 (d7f345585)	MacLean et al., 201056
MSSTATs software suite v3.2.2	Choi et al., 201457

Resource availability

Lead contact

Information regarding this work and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Jennifer E. Van Eyk (Jennifer.VanEyk@cshs.org).

Materials availability

This study did not generate unique reagents.

Experimental model and subject details

Animals

Four-month-old Mat1a^−/− male mice were given vehicle (water, n = 4) or SAMe (in the form of disulfate p-toluene sulfonate dried powder provided by Gnosis SRL (Cairate, Italy), 100 mg/kg/day, n = 4) orally by gavage for one week before sacrificing. Age-matched wild-type male sibling littermates were also treated with vehicle for the same duration (n = 4). Animals were bred, maintained, and cared for as per National Institutes of Health (NIH) guidelines, and protocols were approved by the Institutional Animal Care and Use Committee of Cedars-Sinai Medical Center, Los Angeles, CA. Eight-month-old Mat1a^−/− male mice (in C57Bl/6 background) with increased levels of liver transaminases and fat accumulation on ultrasound were given vehicle (water, n = 6) or SAMe (Abbott, Chicago, IL; 30 mg/kg/day, n = 5) orally by gavage for 8 weeks before sacrificing. Age-matched wild-type male sibling littermates showing normal serum liver transaminases and ultrasound were also treated with vehicle for the same duration (n = 6). Animals were bred and housed in the CIC bioGUNE animal unit, accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC). After harvest, livers from all animals were briefly washed in ice cold PBS containing phosphatase inhibitor cocktail (Roche), snap-frozen in liquid N₂, and stored at −80C until further processing.

Human tissues

De-identified human NAFLD tissues were procured from the Biobank of the Fatty Liver Program at Cedars-Sinai Medical Center (NAFLD repository IRB# Pro00042709). Samples were processed for phospho-proteomics and proteomics analysis as explained in the method details section. Clinical characteristics are listed in Table 2.

Method details

Sample preparation for proteomic analysis

Frozen livers were ground while frozen in a liquid N₂ cooled cryohomogenizer (Retsch). 8M urea and 100 mM TRIS-HCL, pH 8.0 was added the liver power, ultrasonicated (QSonica) at 4°C for 10 minutes with 10 second repeating on/off intervals and centrifuged at 16,000 × g for 10 minutes at 4°C. To the soluble fraction, DTT (15 mM) was added for 1h at 37°C, followed sequentially by iodoacetamide (30 mM) for 30 minutes at room temperature in the dark, diluted to a final concentration of 2M Urea with 100 mM TRIS-HCL, pH 8.0 and Trypsin/Lys-C mix (Promega) at a 1:40 dilution for 16 hours on a shaker at 37°C. Each sample was de-salted using HLB plates (Oasis HLB 30 μm, 5 mg sorbent, Waters) and eluted in 300 μL of 80% ACN, 5% TFA, 1 m glycolic acid.

Titanium dioxide phospho enrichment

Titanium Dioxide (TiO2) phosphorylation affinity enrichment was performed as previously described.⁵⁸ Briefly, 400 μg of digested peptides were incubated in 50 μL titanium dioxide (TiO2) slurry (30 mg/mL, Glygen Corp, Columbia, MD) at room temperature on a shaker for 16 hrs. TiO2 beads were washed twice with 200 μL of 80% ACN, 5% TFA, once with 200 μL of 80% ACN, 0.1% TFA, and eluted in 180 μL of 30% ACN/ 1% NH4OH and neutralized with 200 μL of 10% FA. Samples were then desalted on Oasis HLB μ-elution plates (Waters) and eluted in 80% ACN, 0.1% FA, dried in Speedvac, then resuspended in 0.1% FA for LC–MS/MS analysis.

Quantitation of individual specimen by DIA-MS

Tryptic peptide assay library was created by DDA acquisitions of 5 strong Cation Exchange (SCX) fractions from glycine N-methyltransferase knockout (Gnmt^−/−) mouse livers was we described⁴⁴ was used for DIA analysis. Peak group extraction and FDR analysis was done as we outlined.⁴⁴ Raw intensity data for peptide fragments was extracted from DIA files using the open source openSWATH workflow against the sample specific peptide assay.⁵⁹ Then, retention time prediction was made using the Biognosys iRT Standards spiked into each sample. Target and decoy peptides were then extracted. scored and analyzed using the mProphet algorithm to determine scoring cut-offs consistent with 1% FDR .⁶⁰ Peak group extraction data from each DIA file was combined using the ‘feature alignment’ script, which performs data alignment and modeling analysis across an experimental dataset.⁶¹ Finally, all duplicate peptides were removed from the dataset to ensure that peptide sequences are proteotypic to a given protein in our FASTA database.

DIA-MS data normalization, quantitation, and visualization

The total ion current (TIC) associated with the MS2 signal across the chromatogram was calculated for normalization using in-house software. This ‘MS2 Signal’ of each file was used to adjust the transition intensity of each peptide in a corresponding file. Normalized transition-level data was then processed using the mapDIA software to obtain total protein quantity and perform pair-wise comparisons between groups at the peptide level.⁵⁵ For phospho-proteomics analysis, to obtain the ratio of phosphorylation when compared to total protein quantity (Phospho/Total), the intensity of each unique phospho-peptides was divided by its corresponding protein quantity.

Phospho enrichment database searches and quantification

DDA files were converted to mzXML and searched through the Trans Proteomic Pipeline (TPP) using 3 algorithms, (1) Comet,⁶² (2) X!tandem! Native scoring,⁶³ and (3) X!tandem! K-scoring⁶⁴ against a reviewed, mouse canonical protein sequence database, downloaded from the Uniprot database on January 24^th, 2019, containing 17,002 target proteins and 17,002 randomized decoy proteins. Target-decoy modeling of peptide spectral matches was performed with peptide prophet⁶⁵ and peptides with a probability score of >95% from the entire experimental dataset were imported into Skyline software⁵⁶ for quantification of precursor extracted ion intensities (XICs). Precursor XICs from each experimental file were extracted against the Skyline library, and peptide XICs with isotope dot product scores>0.8 were filtered for final statistical analysis of proteomic differences.⁶⁶ Raw peptide intensities were used to calculate pairwise comparisons between experimental groups using the linear mixed effects model built into the open sources MSSTATs (v3.2.2) software suite.⁵⁷ Peptide abundance differences with a p-value <0.05 were considered significantly different. The Skyline documents containing precursor XICs from each experimental file are available at Panorama. Phospho-peptide quantification files are uploaded on panorama (https://panoramaweb.org/Mat1a_NASH.url, Proteome Exchange ID: PXD022122). Additionally, our recently published dataset containing the phospho-proteomics data from Mat1a^−/− animals with NASH treated with SAMe administration is available at https://panoramaweb.org/Larp1.url.¹⁸

Triglyceride measurement

Liver triglycerides were extracted and quantified using the triglyceride colorimetric assay kit (Cayman Chemicals, MI) following the manufacturer’s protocol. Briefly, tissues and triglyceride standards from the kit were solubilized in a NP-40-substitute buffer. These were further treated with an enzyme mixture containing lipase, glycerol kinase, glycerol oxidase and peroxidase to sequentially generate glycerol, glycerol-3 phosphate, dihydroxyacetone phosphate + H₂0₂. The H₂0₂ reacted with the peroxidase in the presence of 4-aminoantipyrine and N-ethyl-N-(3′-sulfopropyl)-m-anisidine to produce a brilliant purple quinoneimine dye whose absorbance could be measured at a wavelength of 530–550 nm. The absorbance of the triglyceride standards was plotted against their amount in milligrams and the standard curve was used to determine the amount of triglyceride in the tissue samples. Data are represented as mg of triglyceride/gm of tissue.

ALT/AST assay

Serum ALT assay was measured using the ALT activity assay kit according to the manufacturer (Sigma). Serum AST levels were measured by the AST colorimetric assay kit according to the manufacturer’s protocol (Cayman Chemicals).

H&E staining

Paraffin-embedded liver tissues were stained with hematoxylin and eosin (H&E) using the core services provided by the liver histology core of the University of Southern California research center for liver diseases (NIH grant P30 DK048522).

Hepatic peroxisome and mitochondria staining

Liver tissues were fixed in neutral buffered 10% formalin solution (Sigma-Aldrich, HT501128-4L) embedded in paraffin and cut into 5-μm sections. Liver sections were deparaffinized with Histo-Clear I solution (Electron Microscopy Sciences, 64110-01) and hydrated through decreasing concentration of alcohol solutions. For catalase (peroxisomal marker) and VDAC1/2 (mitochondrial marker) staining, sections were unmasked 20 minutes at 600 W in a microwave with citrate buffer, pH 6.0. Samples were then blocked for 10 min with 3% H₂O₂ followed by 30 min with 2.5% normal goat serum. Sections were further incubated overnight at 4°C with primary antibodies (1:100 dilution) (catalase ref: ab16731; VDAC1/2, ab154856, Abcam) followed by 30 minutes of incubation with goat anti-Rabbit IgG Alexa Fluor™ 647 (for catalase staining) or goat anti-Rabbit IgG Alexa Fluor™ 488 (for VDAC1/2 staining) (1:200 dilution) (Thermo Fisher Scientific). Samples were mounted with Fluoromount-G mounting medium with DAPI (Invitrogen). All images were captured using a Carl Zeiss Axioimager D1 fluorescent microscope and quantified for number and size of peroxisome and mitochondria using Image J software.

Quantification and statistical analysis

Normalized transition-level data was processed using the mapDIA software to obtain total peptide and protein quantities, remove data missingness (minimum 50% observations present per experimental group) and perform statistical analysis of pair-wise comparisons between the experimental groups at the protein and peptide level.⁵⁵ Differences between groups were identified by pair-wise comparisons using differential expression analysis based on a Bayesian latent variable model with Markov random field model as previously described.⁵⁵ Proteins/peptides were considered significantly different with the computed Bayesian FDR of less than 0.05.

Published: February 17, 2023

Supplemental information

Table S3. Enriched cellular components and pathways in Mat1a−/− pre-disease and NASH with or without SAMe administration by ClueGo ontology analysis

Raw data of the protein and peptide quantification values from which the ClueGO analysis was derived are uploaded on https://panoramaweb.org/Mat1a_NASH.url, Proteome Exchange ID: PXD022122 related to Figure 2.

Data and code availability

• •Phospho-peptide quantification data have been deposited on panorama (https://panoramaweb.org/Mat1a_NASH.url) and will be shared by the lead contact upon request. The ProteomeXchange ID is listed in the key resources table.
• •Published dataset containing the phospho-proteomics data from Mat1a^−/− animals with NASH treated with SAMe administration is available at https://panoramaweb.org/Larp1.url. (Ramani et al., 2022). The ProteomeXchange ID is listed in the key resources table.
• •This paper does not report original code.