Multi-omics Analysis of Liver Infiltrating Macrophages Following Ethanol Consumption

Alcoholic liver disease (ALD) is a significant health hazard and economic burden affecting approximately 10 million people in the United States. ALD stems from the production of toxic-reactive metabolites, oxidative stress and fat accumulation in hepatocytes which ultimately results in hepatocyte death promoting hepatitis and fibrosis deposition. Monocyte-derived infiltrating Ly6Chi and Ly6Clow macrophages are instrumental in perpetuating and resolving the hepatitis and fibrosis associated with ALD pathogenesis. In the present study we isolated liver infiltrating macrophages from mice on an ethanol diet and subjected them to metabolomic and proteomic analysis to provide a broad assessment of the cellular metabolite and protein differences between infiltrating macrophage phenotypes. We identified numerous differentially regulated metabolites and proteins between Ly6Chi and Ly6Clow macrophages. Bioinformatic analysis for pathway enrichment of the differentially regulated metabolites showed a significant number of metabolites involved in the processes of glycerophospholipid metabolism, arachidonic acid metabolism and phospholipid biosynthesis. From analysis of the infiltrating macrophage proteome, we observed a significant enrichment in the biological processes of antigen presentation, actin polymerization and organization, phagocytosis and apoptotic regulation. The data presented herein could yield exciting new research avenues for the analysis of signaling pathways regulating macrophage polarization in ALD.

www.nature.com/scientificreports www.nature.com/scientificreports/ macrophage subtypes (Table 1). In the lipid positive fraction, we detected 58 differentially regulated metabolites with 39 upregulated and 19 downregulated in the Ly6C low compared to the Ly6C hi macrophages. From the lipid negative fraction, we measured 30 differentially regulated metabolites with 15 upregulated and 15 downregulated in the Ly6C low compared to the Ly6C hi macrophages. In the aqueous fraction, we detected 14 differentially regulated metabolites with 8 being upregulated and 6 downregulated in Ly6C low compared to the Ly6C hi macrophages. Following analysis with Metaboanalyst, we performed Metabolites Biological Role (MBROLE) analysis for pathway enrichment. From the 102 significantly altered metabolites we observed 6 pathways significantly enriched of which glycerophospholipid metabolism, arachidonic acid metabolism and phospholipid biosynthesis were further analyzed for their potential role in regulating macrophage polarization. (Table 2). Ly6C hi and Ly6C low macrophages are significantly enriched for glycerophospholipid metabolism, metabolic pathways, arachidonic acid metabolism, linoleic metabolism and phospholipid biosynthesis with differential regulation of the metabolites involved in each functional pathway (Supplementary Table S1).
Comparative Proteomic Analysis of Ly6C hi and Ly6C low Infiltrating Macrophages Following Ethanol Administration. Following methanol extraction of metabolites, the remaining protein pellet was subjected to protein extraction and tryptic digested for mass spectrometry proteomics analysis. Peptides detected by mass spectrometry were searched in Spectrum Mill to determine the protein identification. We detected 1,304 proteins in Ly6C hi and Ly6C low macrophages with 340 and 214 proteins, respectively, uniquely expressed between macrophage subtypes (Fig. 4A). The 1,304 protein found in the Ly6C hi and Ly6C low macrophages were subjected to DAVID analysis. From the 1,304 proteins analyzed, we observed 429 biological processes of which 105 were unique for Ly6C low and 75 for Ly6C high macrophages (Fig. 4B). Furthermore, we detected 200 molecular functions    www.nature.com/scientificreports www.nature.com/scientificreports/ from the 1,304 proteins of which 23 are unique for Ly6C low and 28 for Ly6C high macrophages (Fig. 4C). The UniProt accession numbers for the common and unique proteins, biological processes and molecular functions are listed in the Supplementary Information Section (Supplementary Tables S3-S5). Protein quantitative analysis of significantly altered proteins was obtained from Mass Profiler Professional and we detected 47 differentially www.nature.com/scientificreports www.nature.com/scientificreports/ regulated proteins between the Ly6C hi and Ly6C low macrophages ( Table 3). The significantly altered proteins between the Ly6C hi and Ly6C low macrophages were analyzed using the DAVID bioinformatics resource and we observed a total of 21 biological processes and 9 molecular functions from DAVID analysis of the protein quantification obtained (Supplementary Table S2). Of the significantly enriched biological processes and molecular functions, immune processes, actin polymerization and organization, phagocytosis, apoptotic processes and antigen presentation were selected for additional literature based analysis in their potential role for regulating macrophage polarization (Table 4).

Discussion
Alcoholic liver disease is a major public health issue and accounts for approximately 48% of liver cirrhosis related deaths 1 . As infiltrating macrophages are known to mediate the pathogenesis of ALD from steatosis to cirrhosis 8-10 , analysis of the altered signaling pathways between the different subsets of these cells in response to ethanol is of the utmost importance in developing treatment options to prevent the progression of ALD or promote the reversal of scar tissue formation in the liver. Macrophages display a remarkable capacity to adapt their phenotype based on tissue micro-environmental cues such as lipid exposure, hypoxia, cytokines, and efferocytosis of apoptotic cells 21,25 . To date, no studies have been conducted providing analysis of the cellular metabolome and proteome of infiltrating liver macrophages isolated from an in vivo model of ALD. While several studies have utilized immortalized mouse macrophages (RAW264.7) for transcriptomic 26 and lipidomic 26-28 analysis following inflammatory stimuli, this study is the first to look at in vivo polarized macrophages in an ALD model, therefore allowing for the natural effects of the tissue microenvironment, such as the gut-liver signaling axis, and ethanol metabolism on regulating liver infiltrating macrophage phenotypes.
It has previously been shown that following phagocytosis of apoptotic hepatocytes, Ly6C hi macrophages differentiate into Ly6C low macrophages which express higher levels of phagocytosis related genes after alcohol intake 21 . In healthy or control diet fed mouse livers, infiltrating macrophages are limited until liver insult elicits the recruitment of Ly6C + monocytes into the liver tissue. 11,21,25 Therefore, the analysis done in this study was focused on the difference between Ly6C hi and Ly6C low macrophages from ethanol fed mice without comparison to control diet fed animals. In our present study, we observed a significant increase in phagocytosis and engulfment related proteins (Table 4). We detected an upregulation of phagocytosis related proteins in Ly6C low macrophages; this is expected as phagocytosis of apoptotic cells induces an anti-inflammatory phenotype 29,30 . Additionally, we saw a significant enrichment in proteins involved in regulating the apoptotic process. Furthermore, we observed a significant enrichment in actin polymerization and cytoskeletal reorganization in Ly6C low macrophages. Alterations in actin contractility, cytoskeletal organization and cellular elongation have been shown to induce macrophages to an anti-inflammatory phenotype as evidenced by increased arginase-1 and YM-1 expression, hallmarks of anti-inflammatory macrophage polarization 31 . Additionally, defects in actin polymerization have been shown to attenuate macrophage phagocytic ability 32 . This suggests further in vivo analysis of actin polymerization and cytoskeletal organization in murine macrophages may elucidate a novel therapeutic strategy in modulating macrophage phenotypes in ALD by affecting macrophage phagocytosis and response to apoptotic stimuli.
Recently Zhang et al. provided a comprehensive analysis profiling lipid species during in vitro differentiation of mouse and human macrophages cell lines. They reported a significant increase in the composition of glycerophospholipid species during macrophage differentiation. Furthermore, they saw a significant increase in the levels of lysophospholipids in anti-inflammatory macrophages compared to pro-inflammatory macrophages suggesting that modulation of glycerophospholipid metabolism could be a vital signaling component in differentiation of liver macrophage phenotypes 33 . In our study, we found a significant enrichment in glycerophospholpid metabolism with differential metabolite regulation between Ly6C hi and Ly6C low macrophages. Additionally, we observed enrichment for arachidonic acid metabolism and phospholipid biosynthesis ( Table 2). In each of the enriched pathways, we detected a massive upregulation in multiple phosphatidylcholine (PC) species in Ly6C low macrophages. PCs has been shown to promote an anti-inflammatory phenotype in macrophages through modulating actin assembly and increasing mycobacterium growth in RAW264.7 and J774 macrophages 34 . Likewise, we observed a substantial upregulation in phosphatidylethanolamine (PE(18:2/18:2) in Ly6C low macrophages. Following stimulation with nonsteroidal anti-inflammatory agents, macrophages have been shown to display an increase in multiple PE species and take on an anti-inflammatory phenotype 35 . Therefore, the observed changes we see in PC and PE species correlate with in vitro studies highlighting the anti-inflammatory properties of PC and PE glycerophospholipid species in modulating macrophage phenotypes. Also of interest in regard to PE(18:2/18:2) is the linonleic acid (18:2) constituents present at the sn-1 and sn-2 positions, as linoleic acid has been shown to promote an anti-inflammatory phenotype in macrophages 36 . These results suggest the involvement of phospholipase A 2 (PLA 2 ) in regulating macrophage polarization in ALD. PLA 2 is involved in the hydrolysis www.nature.com/scientificreports www.nature.com/scientificreports/ of sn-2 fatty acids from membrane glycerophospholipids yielding a free fatty acid, arachidonic acid, and a lysophospholipid 37 . The functions of PLA 2 in modulating the inflammatory response have been well established in a variety of inflammatory contexts 38 www.nature.com/scientificreports www.nature.com/scientificreports/ metabolism and cyclooxygenase-2 (COX-2) mediated prostaglandin production, elicited an anti-inflammatory and protective response in ALD 45 . In addition to COX-2 mediated arachidonic acid metabolism and prostaglandin synthesis, arachidonic acid can be metabolized via cytochrome P450 epoxygenase mediated pathway to generate epoxyeicosatrienoic acids (EETs) 46 . We found a significant increase in EETs in the Ly6C low phenotype. Endogenous EETs have been shown to regulate the ability of in vitro THP-1 monocytes to differentiate into pro-inflammatory macrophages in response to pro-inflammatory stimuli (lipopolysaccharide (LPS) and interferon γ (IFNγ) as well as preventing differentiation under anti-inflammatory stimuli (IL-4) 46 . Additionally, it has been shown that the immunomodulatory effect of EETs on inducing pro-inflammatory macrophage differentiation was facilitated through attenuation of NF-κB signaling 47 . Finally, studies have shown that eicosatrienoic acid inhibits LPS induced inflammatory gene expression in macrophages 48 . We detected an upregulation of eicosatrienoic acid metabolites in the anti-inflammatory, Ly6C low macrophages after alcohol consumption. These studies coupled with the observed increase in arachidonic acid, glycerophospolipid metabolism and phospholipid biosynthesis as well as increased calcium ion binding suggest future investigation of the role of calcium dependent and independent PLA 2 activity for therapeutic targeting of macrophage polarization in ALD.
The present study provides a framework for future studies utilizing multi-omics approaches for analyzing signaling difference between pro-and anti-inflammatory macrophages isolated from ALD mouse models. We detected a number of metabolic and protein mediated pathways that were significantly altered between the two macrophage subtypes, validating a number of in vitro studies analyzing the lipid, metabolite, and protein profile of polarized macrophages [26][27][28]33,48 . While the present study utilized an ALD model in which the degree of inflammation is not as evident histopathologically as more aggressive models, such as the NIAAA model, it allowed for the sufficient isolation of infiltrating liver macrophages not normally present in the healthy liver. We identified a number of metabolic pathways significantly altered due to the early onset of alcohol-induced hepatic inflammation (arachidonic acid metabolism, glycerophospholipid metabolism and phospholipid biosynthesis), which suggests that PLA 2 enzymes play a critical role in modulating macrophage phenotypes. To explore the impact of PLA 2 on ALD, future studies could utilize whole body PLA 2 knockout mice or known PLA 2 pharmacological inhibitors to elucidate the impact of PLA 2 on macrophage polarization in ALD models. Overall, the data presented here justifies a further need to investigate numerous signaling mechanisms implicated in the modulation of macrophage phenotypes during ALD.

Materials and Methods
Animal Model. Female C57BL/6 J mice (The Jackson Laboratory, Bar Harbor, ME, USA) (n = 30) were maintained under pathogen-free conditions in the Center for Laboratory Animal Care at the University of Colorado Anschutz Medical Campus (Aurora, CO, USA). All experiments were performed using an Institutional Animal Care and Use Committee (IACUC) approved protocol and in accordance to the guidelines of the IACUC at the University of Colorado Anschutz Medical Campus. To elicit infiltrating macrophage recruitment to the liver, mice were fed an ethanol-containing Lieber-Decarli liquid diet (Bio-Serv, Flemington, NJ, USA). Ethanol content was introduced gradually by increasing 1.6% (v/v) every 2 days until 5%. All mice were then fed the liquid diet containing 5% ethanol for 4 weeks, as described previously 49,50 .

Isolation of Liver Non-Parenchymal Cells (NPCs).
Liver NPCs were isolated following a previously described method 51 . Briefly, a 20-G catheter was put through the mouse superior vena cava, the inferior vena cava was clamped, and the portal vein cut. The liver was perfused with Hank's balanced salt solution (HBSS), followed by a digestion buffer [1 × HBSS, supplemented with 0.04% collagenase (type IV; Sigma, St. Louis, MO, USA), 1.25 mM CaCl 2 , 4 mM MgSO 4 , and 10 mM HEPES]. After digestion, the liver was disrupted in ACD solution (1 × HBSS, supplemented with 0.5% FBS, 0.6% citrate-dextrose solution, and 10 mM HEPES). Single cells were passed through a 100-μm cell strainer, and the cells were fractionated using 30% (w/v) Nycodenz (Axis-Shield PoC AS, Oslo, Norway) at 1.155 g/mL to yield liver NPCs and further purified using 30% Percoll (Sigma) at 1.04 g/ mL.

Metabolomics Sample Preparation and Analysis.
Cell pellets from different sort dates were combined in order to get 3 technical replicates of approximately 400-500 K cells per sample type (Ly6C hi and Ly6C low ). Extractions were performed using volumes of 70% MeOH/water and 100% MeOH based on cell numbers. Cold methanol was used to precipitate proteins prior to liquid-liquid extraction of metabolites. Proteins pellets were saved for future proteomics analysis. Liquid-liquid extraction was performed on the supernatant using water and methyl tert-butyl ether (MTBE).The aqueous and lipid fractions were retained for analysis. Lipid fractions were analyzed using SB-C18 HPLC analytical column in positive and negative ionization mode on the Agilent 6560 IM-QTOF (in QTOF mode only). Aqueous fractions were analyzed using a HILIC column in positive ionization mode on the Agilent 6560 IM-QTOF (in QTOF mode only). A pooled sample was used as instrument QCs to monitor the entire instrument analysis. Initial data QC, peak threshold evaluation, retention time variation, and charge carrier evaluation was performed in Agilent MassHunter Qualitative Analysis, version B.07.00. Data extraction was performed in MassHunter Profinder, version B.08.00. Differential Analysis was (2019) 9 www.nature.com/scientificreports www.nature.com/scientificreports/ performed in Agilent Mass Profiler Professional (MPP), version 14.5. Compound annotation (database searches and molecular formula generation) was performed in MassHunter ID Browser software, version 14.5. Raw MS data were checked for quality and reproducibility. Appropriate spectral and chromatogram peak height thresholds were determined by careful examination of the raw data. Appropriate charge carriers to be allowed during data extraction were determined after preliminary extraction on selected samples. The Agilent "recursive workflow" was used to prepare data. This workflow includes the following steps: 1) untargeted extraction using the Find-by-Molecular Feature algorithm, 2) mass and time alignment of extracted compounds, 3) targeted extraction using the Find-by-Ion algorithm (using the list of ions prepared in step 1), 4) final mass and time alignment of extracted compounds.

Proteomics Sample Preparation, nHPLC-MS and nHPLC-MS/MS Analysis. Following metha-
nol extraction for metabolomics, the remaining cell pellets from each technical replicate were processed using the PreOmics iST 8x Kit (Cat # 00001) following the included protocol. Digested macrophage samples were loaded onto a 2 cm PepMAP 100, nanoviper trapping column and chromatographically resolved on-line using a 0.075 × 250 mm, 2.0 µm Acclaim PepMap RSLC reverse phase nano column (Thermo Scientific) using a 1290 Infinity II LC system equipped with a nanoadapter (Agilent). Mobile phases consisted of water + 0.1% formic acid (A) and 90% aq. acetonitrile + 0.1% formic acid (B). Samples were loaded onto the trapping column at 3.2 μL/min for 2.5 minutes at initial condition before being chromatographically separated at an effective flow rate of 345 nl/ min using a gradient of 3-8.5% B over 4.0 minutes, 8.5-26% B over 48.5 minutes, and 26-35% over 7.5 minutes for a total 60 minute gradient. The gradient method was followed by a column wash at 70% B for 5 minutes. For nHPLC-MS, data was collected with a 6550 QTOF equipped with a nano source (Agilent) operated in MS mode. For nHPLC-MS/MS, data was collected with a 6550 QTOF equipped with a nano source (Agilent) operated using Data Dependent Acquisition CID Auto MS/MS. The capillary voltage, drying gas flow, and drying gas temperature were set to 1300 V, 11.0 l/min, and 200 C, respectively. Data was collected in positive ion polarity over mass ranges 290-1700 m/z at a scan rate of 1.5 spectra/s. MS/MS scans were collected over mass ranges 50-1700 m/z at a scan rate of 3 spectra/s. Singly charged species were excluded from being selected during MS/MS acquisition. Following data acquisition in MS/MS mode, sample data was searched in SpectrumMill to identify proteins.

DAVID Bioinformatics Analysis. Functional pathway enrichment of significantly altered proteins between
Ly6C hi and Ly6C low macrophage population was analyzed using the Database for Annotation, Visualization and Integrated Discovery (DAVID) Bioinformatics Resource 6.8. For pathway enrichment, significantly altered proteins were compared to the whole mouse background. Pathways were considered significant with at least 3 proteins involved, a fold enrichment >2, and a p < 0.05. Statistical Analysis. Statistical analysis of significantly altered metabolites and proteins was determined using Mass Profiler Professional Software. For Metaboanalyst, significantly altered metabolites were determined  Table 4. DAVID functional pathway enrichment of significantly altered proteins between Ly6C hi and Ly6C low macrophages. (n = 3 in each experiment). Pathway enrichment was considered significant with a DAVID calculated t-test p < 0.05 when comparing Ly6C hi and Ly6C low .
www.nature.com/scientificreports www.nature.com/scientificreports/ based of the difference in peak height intensity between the analyze metabolites with a p < 0.05. For MBROLE analysis for metabolite functional pathway enrichment, pathways were considered significant with a p < 0.05. For DAVID pathway enrichment, significantly altered proteins were compared to the whole mouse background. Pathways were considered significant with at least 3 proteins involved, a fold enrichment >2, and a p < 0.05.