Transcriptome of monocytes from liver, brain and bone marrow reveals organ-specific features in aging and alcohol misuse

Abstract

Background & Aims

Aging and alcohol misuse independently alter monocyte (MO) and macrophage (MØ) function, leading to impaired antimicrobial responses. However, how alcohol misuse contributes to impaired MO/MØ function during aging remains unclear.

Methods

We compared the transcriptomes of MOs and MØs from alcohol-modulated niches (liver, brain, and bone marrow [BM]) in young (3-month-old) and old (20–24-month-old) female C57BL/6N mice (n = 4–6 per group). Statistical significance was determined using two-way ANOVA.

Results

MO/MØ transcriptomes showed unique organ-specific responses to aging and alcohol. Aging elicited a common deregulation of pathogen-responsive pathways, while alcohol misuse commonly inhibited IFN signaling in the aged populations. Our studies on intercellular communication using ligand-receptor interactions revealed that BM MOs were the least communicative and liver MØs were the most communicative. Alcohol misuse specifically increased MO/MØ communication in aging. We also identified and validated specific pathways driving inter-organ MO/MØ crosstalk in alcohol misuse during aging, including APOE-TREM2 signaling from the liver to microglia and the NRXN2 and SPP1 pathways.

Conclusion

Our results provide a unique insight into the heterogeneity of the MO/MØ transcriptome and define the inter-organ crosstalk between BM, liver, and brain during aging and alcohol misuse.

Impact and implications

Aging and alcohol misuse are linked to immune dysfunction, systemic inflammation, and altered innate immune responses. Here, we examined monocyte/macrophage responses in the liver, brain, and bone marrow of young and aged mice under alcohol exposure at the transcriptomic level. We observed that aging and alcohol predominantly elicited organ-specific changes in gene expression, with minimal overlap between the monocyte/macrophage populations across different tissues. However, aging commonly upregulated pathogen response pathways while alcohol misuse inhibited interferon signaling. We also assessed cell-cell communication by analyzing ligand-receptor expression in the different monocyte/macrophage populations and identified candidate molecules (APOE, TREM2, NRXN2, SPP1) from the top pathways guiding inter-organ signaling specifically in aging and alcohol misuse. Our findings have generated a unique repository and provide novel insights on how aging and alcohol impact tissue-specific monocytes/macrophages and their crosstalk.

Graphical abstract

Highlights

• •Monocytes/macrophages from aged-liver, brain and bone marrow show a predominant organ-specific change in differentially expressed genes with overlap of 7 genes establishing an aging monocytes/macrophage signature.
• •Monocytes/macrophages from aged-liver, brain and bone marrow show common deregulation of viral response pathways, NOD-like receptor signaling pathways, JAK-STAT signaling and TNF-NFκB signaling.
• •Alcohol misuse commonly inhibits IFN signaling in the monocytes from liver, brain and bone marrow.
• •Alcohol misuse and aging both modulated inter-cellular crosstalk between the monocytes/macrophages from the three organs, with alcohol increasing the expression of APOE and NRXN2 in liver MO/MØ and TREM2 and SPP1 in microglia.

Introduction

According to the US Department of Health and Human Services Report from 2022, adults aged 65 and older account for 17.3% (58 million) of the US population, and this proportion is projected to increase to 22% by 2040. Age is the most common risk factor for chronic diseases,¹ and the immune system becomes less efficient in fighting pathogens,² making elderly people more prone to contracting infections and developing sepsis.³ Inflammaging, defined as the basal low-grade pro-inflammatory state in aged organs, is a major contributor to immune dysfunction.⁴ Moreover, the immune system influences the rate of aging of multiple organs/systems, including cardiovascular and musculoskeletal systems, raising a person’s mortality risk beyond existing chronic disease, age, or sex.⁵

Alcohol misuse is increasing, particularly among middle-aged and older adults.⁶ Chronic or heavy alcohol consumption is known to impair the maturation and function of cells of the innate immune system^7,8 and increase the risk of infections and associated mortality[9], [10], [11], [12] by disrupting gut permeability and allowing bacterial products to transfer to the systemic circulation. Additionally, aged individuals are more susceptible to the toxicity induced by alcohol at a multi-organ level,¹³ including injuries, dehydration, memory loss, liver disease, interaction with medications, and overall mortality.¹³ This sensitivity can be explained – at least partially - by a decline in the activity of alcohol-metabolizing enzymes, a decline in lean body mass, and impaired balance after drinking, increasing the likelihood of falls and accidents with age.¹⁴ In the context of alcohol misuse, the liver is considered the main site for alcohol metabolism and detoxification, while in the brain, astrocytes have been shown to metabolize alcohol,¹⁵ and this may contribute to addiction and alcohol-related misbehaviors.

Monocytes (MOs) and tissue macrophages (MØs) in the liver (Kupffer cells) and brain (microglia) play a pivotal role in both immune defense against pathogens as well as in alcohol-associated inflammation.¹⁶ Both microglia and Kupffer cells originate embryonically from the yolk sac.^17,18 Contrarily, infiltrating MØs are derived from circulating MOs that originate from hematopoietic progenitors in the bone marrow (BM) and are then recruited to sites of injury. During aging, senescent hematopoietic progenitors are skewed towards increasing myeloid progenitors (giving rise to MO/MØ populations) at the expense of lymphoid-derived cells,⁴ contributing to imbalanced immune responses during aging.

To date, little is known about immune inter-organ crosstalk in alcohol-affected organs in the context of aging and/or alcohol misuse. Importantly, the multi-organ aging network suggests that age-related changes in a given organ influence the aging outcomes in other organs.⁵

Thus, in the current manuscript, we dissected MO/MØ phenotypes and their communication potential (probability of ligand-receptor interaction) at the main sites for alcohol metabolism (liver), behavioral response to alcohol (brain), and MO source (BM) during aging, alcohol misuse, and their combination. We also validated the top communication pathway candidates (APOE, TREM2, NRXN2, and SPP1) in liver and brain sections from these mice. Importantly, we showed that alcohol misuse worsened inflammaging, characterized by circulating levels of IL1β and decreased phagocytosis signatures in old mice receiving alcohol. We created the first data repository and compared the transcriptomic changes in MO/MØ populations from the liver, brain, and BM in a mouse model of alcohol misuse and aging.

Materials and methods

Animals

Old (20-24-month old) or young (3-month-old) female C57BL/6N mice were provided by the National Institute of Aging or purchased from The Charles River Laboratory (Cambridge, MA). Mice were housed in a specific pathogen-free mouse facility at Beth Israel Deaconess Medical Center, and all animal handling was performed in compliance with institutional guidelines. Additional procedures were approved by the Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee (protocol #010-2019 and #030-2022).

Alcohol binge administration

Acute administration of ethanol (3.5 g/kg) by oral gavage was adapted from.⁴⁹ Briefly, a 47.3% (vol/vol) ethanol solution was prepared from pure alcohol (1000002000, Pharmco) in water. Volume administration to each mice was calculated as follows: gavage volume (μl) of 47.3% (vol/vol) solution for each mouse = mouse body weight in grams × 15.

Microglia isolation

Cells were isolated as described previously.⁵⁰ Briefly, one half of the brains (devoid of hippocampus, cerebellum) were homogenized using a mortar and pestle in ice-cold HBSS and filtered through 70 μm filter. Cells were pelleted and resuspended in 37% Percoll gradient. Cells were spun at 2,000 RPM for 25 min with breaks off, and the pelleted cells were resuspended in 2% FBS in PBS for fluorescence-based staining and subsequent sorting.

Liver mononuclear cell isolation

Livers were perfused with collagenase and mechanically disrupted to obtain a multicellular suspension. The slurry was passed through a 70 μm filter and spun at 250-300 RPM. The top layer of non-parenchymal cells was collected and spun at 1,800 RPM. The pellet was resuspended in 40% Percoll and layered on top on 70% Percoll (17-5445-01, Cytiva). The Percoll gradient was centrifuged at 2,000 RPM for 30 min without breaks. Cells at the interface were harvested and washed with PBS and stored in 10% DMSO 90% FBS at -80° for fluorescence-based staining and subsequent sorting.

BM cell isolation

Tibia and femur from mice were used for BM isolations. Both ends of the bone were trimmed to expose the interior marrow shaft. The bone pieces were placed in 0.2 ml Eppendorf tubes and centrifuged at 8,000 G for 10 min in 2% FBS in PBS. Cells were treated with red blood cell lysis buffer and washed with 2% FBS in PBS.⁵¹ Cells were then filtered through a 40 μm filter, pelleted, and stored in 10% DMSO 90% FBS at -80° for fluorescence-based staining and subsequent sorting.

Cell sorting

Microglia, liver mononuclear fraction and bone marrow multicellular suspensions were incubated with a cocktail of antibodies (1:100 dilution) for 30 min on ice: APC/Cyanine7 anti-mouse CD45 Antibody (Biolegend, 103116, clone 30-F11), Brilliant Violet® 711 Anti-CD11b Rat Monoclonal Antibody (Biolegend, Cat#101242, clone: M1/70), Alexa Fluor® 700Anti-Ly-6C Rat Monoclonal Antibody (Biolegend, Cat#128024, clone: HK1.4), Brilliant Violet 785™ anti-mouse Ly-6G Antibody (Biolegend, Cat#127645, clone: 1A8), PE anti-mouse F4/80 Antibody (Biolegend, Cat#111703, clone: W20065D), and Zombie NIR™ Fixable Viability Kit (Biolegend, Cat#423106). After 30 min, all samples were incubated with BD Phosflow™ Lyse/Fix Buffer (BD Biosciences, 558049) following the manufacturer’s protocol. Following lyse/Fix, cells were washed twice with 1 × PBS containing 2% FBS, and resuspended in FACS buffer containing 2% FBS in PBS. Samples were run in BD FACSymphony™ S6 (BD Biosciences), and data analysis was done using Flowjo version 8.8.7 software.

Flow cytometry

We isolated and sorted four distinct MO/MØ populations as follows: 1) microglia; 2) liver MO; 3) liver MØ, including both Kupffer cells and infiltrating MØ; and 4) BM MOs from our experimental groups and profiled those by Smart RNA sequencing (Fig. S2A). After gating the cells for singlets and viability, the following surface markers were used to define MO/MØ populations: CD45^{+(intermediate)}CD11b⁺ for microglia, CD45⁺CD11b⁺Ly6C⁺Ly6G^- for liver and BM MOs, and CD45⁺CD11b⁺Ly6C⁺Ly6G^-F4/80⁺ for liver MØs (Fig. S2B). Data retrieved from the Smart RNA-seq analysis passed all quality control tests, including sequence quality scores (Fig. S2C) and mitochondrial and ribosomal transcript content of less than 4% (Fig. S2D–E), indicating high-quality, viable cells. On average, 13,000 to 17,000 transcripts were identified in each sample (Fig. S2F) with homogeneous distribution across the four experimental conditions (Fig. S2G). PC1 segregated microglia samples from the three remaining non-brain populations, while PC2 separated BM and liver populations (Fig. S2H). Interestingly, cell cycle scores for S (synthesis) and G2/M (division) phases correlated with the known life span for each cell type, low for the comparatively longer-lived microglia and higher in the rapidly replenishing BM MOs (Fig. S2I). Antibodies used for staining were from Biolegend as referenced in the supplementary CTAT table.

Immunofluorescence in liver and brain sections

For brain sections, fixed tissues were cryopreserved in 30% sucrose until submerged, then sectioned at 25 μm using a Leica cryostat and stored in 0.05% sodium azide in PBS at 4 °C. Liver sections were prepared from paraffin-embedded tissue, deparaffinized, rehydrated, and processed similarly. For both tissues, sections were permeabilized (0.5% Triton X-100), then blocked (brain: 10% horse serum; liver: 5% goat serum). Primary antibodies were diluted in blocking buffer and incubated overnight at 4 °C. For brain sections, the primary antibody was washed, and the sections were incubated with fluorophore-conjugated secondary antibodies for 1 h at room temperature. Liver sections were incubated with primary antibodies conjugated to a fluorophore. Fluorophore conjugation was carried out as per the Fluorophore conjugation kit protocol (ab236553, ab236553) and counterstained with DAPI (1:1,000, 5 min), washed, and imaged using a Zeiss LSM-700 confocal microscope (63 × magnification). Liver slides were mounted in ProLong™ Gold Antifade Mountant and imaged using a Zeiss LSM 880 upright microscope. IBA1 immunostaining was used in brain sections for morphometric analysis. The antibodies used were as per the supplementary CTAT table.

SMART RNA sequencing

One thousand sorted MOs/MØs were suspended in 1% beta-mercaptoethanol in TCL buffer and sequenced using the Smart-Seq2 platform at the Broad Institute. Briefly, the raw sequencing reads were quality-checked, and data were pre-processed with Cutadapt (v2.5) for adapter removal. Gene expression quantification was performed by aligning against the GRCm38 genome using STAR (v2.7.3a).

Data analysis

Raw output files were analyzed with the Galaxy platform.⁵² Briefly, quality control of the raw sequence data was performed using the FastQC package. Data was then aligned to the mouse genome using the bowtie2 package, and raw counts were obtained with the featureCounts package. All samples passed the following quality criteria: number of genes >300, mitochondrial RNA <30%, ribosomal RNA <30%. Raw counts were normalized using the DESeq2 package for R,⁵³ and differential expression between experimental conditions was evaluated with the same package. Genes were considered differentially expressed when the fold change was >1.5 and the p value <0.05. p values were calculated with the computeCommunProb() function of the package. Pathway analyses for each cell type were run with the pathfinder 2.4.1⁵⁴ and gprofiler2⁵⁵ packages for R, while ligand-receptor interactions between different cell types were predicted with the cellChat 2.1.2²⁰ package for R. Pathway analysis was also performed using Ingenuity Pathway Analysis.

Immunoblotting

Murine liver tissue was lysed in RIPA buffer (BP-115-500, Boston BioProducts) containing protease inhibitor cocktail (5872S, Cell Signaling Technology). Equal amounts of total protein lysates (20-50 μg) were run on SDS-PAGE and transferred to a nitrocellulose membrane. Following blocking (5% BSA in Tris-buffer containing 1% Triton X-100), blots were probed with primary and fluorescent secondary antibodies (IRDye 800 and 700 CW, LI-COR Biosciences) and developed using a LI-COR reader. The antibodies are listed in the supplementary CTAT table.

Blood alcohol content

Levels of blood alcohol (LS-K299, LifeSpan Biosciences), acetaldehyde (Millipore Sigma, MAK434-1 KT), and acetate (Millipore Sigma, MAK474) were measured in mouse serum using colorimetric commercial assays per manufacturer’s instructions. Blood was collected from the tail vein in citrate-coated tubes. Serum was obtained by centrifuging the blood for 10 min at 10K RPM, 4 °C.

mRNA isolation and quantitative real-time PCR

Total RNA was extracted with RNeasy® Mini Kit (74106, Qiagen) per manufacturer's instructions. RNA yield was quantified by the Nanodrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). Up to 1 μg RNA was reverse transcribed using a High-Capacity cDNA Reverse Transcription Kit (1708891BUN, Bio-Rad) in a cDNA Master cycler X50s (Eppendorf). qPCR was performed in a CFX96 Real-Time PCR System (Bio-Rad), using Sybr Green primers and qPCR Master Mix (1725124, Applied Biosystems). Primers used included: F- AGGCTGTCAAGGAGGTGCTACT, R- AAAACCTCCGCACGTCCTTCCA for Cyp2e1 and F- AGGTCGGTGTGAACGGATTTG and R- TGTAGACCATGTAGTTGAGGTCA for Gapdh (control).

Serum cytokine measurements

Cytokine levels were measured in the mouse serum by Eve Technologies (Alberta, Canada) using their proprietary Multiplexing LASER bead technology platform.

Mice behavior

Mice were placed in the dimly lit experimental room for acclimation for 30 min.

An open field test was performed as previously described⁵⁶ to assess locomotion. Mice were individually placed in the center of the open field arena (40 × 40 × 40 cm) facing away from the observer and allowed to freely explore for 5 min. Distance moved and the movement velocity were recorded by a video tracking system (EthoVision XT, Noldus). The arena was thoroughly cleaned with a non-alcohol-based product between trials.

Elevated plus maze: Mice were placed in the central zone facing an open arm, opposite the experimenter, and allowed to explore the arena for 5 min. Anxiety was assessed by the total time spent in the open arms. Each arm measured 20 cm.

Statistical analysis

Appropriate statistical tests that provide adequate power for a given sample size were chosen. Data are presented as mean ± standard deviation. Multiple group analysis was performed using a two-way ANOVA followed by the Tukey post hoc analysis to assess interaction. Differences were considered significant at p <0.05. Statistical analyses were performed using GraphPad Prism 8.0.2 software (GraphPad Software, Inc., San Diego, CA, USA).

Gene set-enrichment analysis

The gene set-enrichment analysis (GSEA) was performed in R using the clusterProfiler package. Briefly, the full gene list from the RNA-seq analysis was sorted according to the "Stat" parameter from the DESeq analysis. This sorted gene list, together with the gene set of interest, was fed to the GSEA function of the package to obtain the enrichment scores and p values. The results were plotted using the gseaplot2() function of the same package.

Results

Aged mice show delayed alcohol clearance in the blood and exacerbated alcohol-induced intoxication

To understand changes in the MO/MØ landscape under the effect of aging and/or alcohol misuse, we chose a repeated alcohol binge paradigm.¹⁹ Briefly, young (3 months) and old (20-24 months) female mice received 10 alcohol binges (3.5 g/kg) or water binges every other day and were euthanized 18 h after the last alcohol binge (Fig. 1A).

Fig. 1.

Aged mice show delayed alcohol clearance in the blood and exacerbated alcohol-induced intoxication. (A) Schematic representation of experimental design. (B) Blood alcohol content dynamics in the serum of the mice after the first binge. (C) Blood alcohol content dynamics in the serum of the mice after the tenth binge. (D) Blood acetaldehyde content dynamics in the serum of the mice after the tenth binge. (E) The blood acetate content dynamics in the serum of the mice after the tenth binge. ∗Refers to significance comparing the old-alcohol group to the old-water group, and ^#refers to significant changes comparing the old-alcohol group with the young-alcohol group. (F) Alcohol intoxication in mice 2 h after receiving the first alcohol binge. (G) Alcohol intoxication in mice 2 h after receiving the tenth alcohol binge. (H) Anxiety-like behavior in mice 18 h after the tenth binge. (I) Relative Cyp2e1 mRNA expression in mouse liver normalized to Gapdh expression. (J) Relative CYP2E1 protein expression in mouse liver normalized to GAPDH expression. Data are expressed as mean ± standard deviation (n ≥4 per group). Statistical significance was determined using two-way ANOVA; ∗p <0.05, ∗∗p <0.001, ∗∗∗p <0.0001, ∗∗∗∗p <0.00001, vs. water-treated controls ^#p <0.05 vs. young mice receiving alcohol. BAC: blood alcohol concentration.

We first explored blood alcohol concentrations (BAC) of the mice after the first and the last (tenth) binge. BAC was significantly higher in the old compared to the young mice for the first 4 h after the alcohol binges (Fig. 1B,C). Furthermore, after the tenth binge, BAC from old mice took twice as long (9 h) to return to baseline compared to young mice (4 h) (Fig. 1B), indicating a delayed alcohol clearance in the old condition (Fig. 1C). Interestingly, BAC 2 h after the last alcohol binge (48.7 mM in young and 97.4 mM in old) was significantly higher compared to the first binge (13 mM in young and 40 mM in old) in both alcohol-treated groups, indicating increased BAC upon repeated alcohol binging regardless of age (Fig. 1B,C). To further dissect alcohol metabolism in our animal model, we measured the circulating levels of acetaldehyde and acetate at 2, 4, and 9 h after the tenth alcohol binge (Fig. 1D,E). At 2 and 4 h after the alcohol binge, acetaldehyde levels were significantly higher in both alcohol-treated groups, regardless of age (Fig. 1D). Interestingly, 9 h after the alcohol binge, acetaldehyde levels in aged mice receiving alcohol remained significantly higher than in water-treated animals and young mice receiving alcohol (Fig. 1D). Blood acetaldehyde levels for the young-alcohol group normalized to young-water levels at 9 h (Fig. 1D). Analysis of acetate levels in the serum showed significantly higher blood acetate levels for the old-alcohol group at all three time points (Fig. 1E). Moreover, at 9 h after the last alcohol binge, acetate levels were significantly higher in the aged group than in the young group (Fig. 1E).

Next, we assessed alcohol-induced behavioral changes (intoxication) in the open field test. Intoxication 2 h after the first alcohol binge was comparable in young and old mice, both showing significantly decreased distance traveled compared to water binge controls (Fig. 1F). However, after the tenth binge, old animals showed significantly more intoxication compared to the young-alcohol group (Fig. 1G). Repeated alcohol administration resulted in anxiety-like behavior (less exploratory behavior) assessed by the elevated plus maze test 18 h after the last alcohol binge (Fig. 1H) in the aged mice. To explore potential reasons for the higher BAC in old mice, we tested the protein expression of the alcohol-metabolizing enzymes – ADH1, ALDH2, and CYP2E1, of which CYP2E1 levels were significantly lower in old compared to young livers at baseline. Repeated alcohol administration significantly reduced CYP2E1 mRNA and protein levels in the young, but not old, livers (Figs 1I,J and S1).

MO/MØ populations show niche-specific transcriptomic profiles in response to aging

In order to investigate the MO/MØ response in this model, we flow-sorted MO/MØ populations from the liver (MO and infiltrating MØ + Kupffer cells), brain (microglia), and BM MO for transcriptomic profiling using the SMART-Seq2 platform (Fig. S2A and B). We first evaluated the differentially expressed genes (DEGs) in response to age (Fig. 2A,B). Throughout the manuscript, we use cut-offs of log₂fold-change ≥1.5 and p <0.05 (complete list in Table S1) for the DEGs. We observed the maximum number of distinct DEGs in microglia (DEG count 674) and in liver MOs (DEG count 633). Liver MØs showed substantially lower DEGs (DEG count 253) with the least DEGs observed in BM MOs (DEG count 167), suggesting that microglia and liver MOs are more sensitive to age-related changes (Fig. 2B). Also, microglia were the most segregated cell type in our analysis based on principal component analysis (Fig. 2C). The number of DEGs induced by aging in these cell types is roughly associated with the cell cycle scores and life span of these cell types (Fig. S2I).

Fig. 2.

MO/MØ populations show niche-specific transcriptomic profiles in response to aging. (A) Volcano plots depicting the significant DEGs (fold change >1.5 and p <0.05) for each cell type. (B) Upset plot representing the number of DEGs in response to age in the different cell types. (C) Principal component analysis for young and aged water groups for each cell type. (D) Dot plot with the relative expression of the seven genes in the MO aging signature. (E) Top 10 pathways expressed in microglia, liver MOs, and BM MOs in response to aging. (F) Dot plots representing relative expression of key genes related to 1) viral response, 2) NOD-like receptor signaling pathway, 3) JAK-STAT signaling pathway, and 4-TNF and NF-κB signaling pathways. Transcriptomic analysis was performed in n = 4-6 mice per experimental condition and cell type. BM, bone marrow; DEGs, differentially expressed genes; MOs, monocytes; MØs, macrophages.

Among these four cell types, we identified seven DEGs that were commonly dysregulated in response to aging (Fig. 2D). These included upregulated genes Cd22, Ide, Lrmda, and downregulated genes Dynlt1b, Wdfy1, Fkbp5, and Serinc3 as an aging signature cluster (Fig. 2D).

Next, we performed pathway analysis on the age-specific DEGs using the pathfindR 2.4.1 package in R. Interestingly, we found commonly dysregulated pathways across the different MO/MØ populations with an enrichment of genes related to viral response pathways, NOD-like receptor signaling pathways, JAK-STAT signaling, and TNF-NFκB signaling (Fig. 2E,F). We found that some of the genes that constitute these pathways had different trends and were up- or downregulated among the four cell types (Fig. 2F). As a corollary to the common pathways, we also identified unique pathways specific to each cell type; for example, microglia showed downregulation of relaxin signaling while liver MOs showed upregulation of fructose and mannose metabolism (Fig. 2E).

Alcohol misuse inhibits interferon signaling across MO/MØ populations

Having determined the effects of aging on MOs/MØs from the brain, liver, and BM, we next investigated the effects of alcohol misuse. Based on the principal component analysis, aging had a more prominent effect than alcohol administration in segregating sample clusters in our model (Fig. 3A). We first looked at the effect of alcohol in the young population (Fig. 3B). With alcohol, microglia had the maximum number of DEGs (209), followed by liver MOs (182) while liver MØs and BM MOs were minimally affected with only 14 and 13 DEGs, respectively, in the young group (Fig.3B) (complete list of DEGs are depicted in Table S2). These data showed that the transcriptome of microglia and liver MOs is more responsive to alcohol compared to liver MØs and BM MOs. Remarkably, there were no common DEGs between the four cell types in the young alcohol-treated samples (Fig. 3B), further suggesting niche-specific changes in response to alcohol.

Fig. 3.

Alcohol misuse inhibits interferon signaling across MO/MØ populations. (A) Principal component analysis for the four experimental groups (young-water, young-alcohol, old-water, and old-alcohol) in each cell type. (B) Upset plot representing the number of differentially expressed genes in young mice exposed to alcohol in the different cell types. (C) Upset plot representing the number of differentially expressed genes in old mice exposed to alcohol in the different cell types. (D-F) Deregulated interferon pathway in microglia, liver MOs, and BM MOs in old mice receiving alcohol. Transcriptomic analysis was performed in n = 4-6 mice per experimental condition and cell type. BM, bone marrow; MOs, monocytes; MØs, macrophages.

With aging (Fig. 3C), BM MOs were most affected by alcohol treatment with 180 DEGs, followed by microglia (162) and liver MOs (135), showing increased susceptibility of old MOs/MØs to alcohol misuse (complete list in Table S3). The liver MØs only had one DEG, Cxcl13.

Next, we performed Ingenuity Pathway Analysis to further understand the main pathways characterizing the MO/MØ populations in response to alcohol. Activation of B cell receptor and inhibition of TGFβ pathways were commonly affected by alcohol in young and old microglia (Fig. S3A and B) while IL-12 was exclusively inhibited in young microglia exposed to alcohol (Fig. S3C). Regarding liver MOs (Fig. S3D), pathways leading to the activation/production of IL-4, IL-5, IL-1β and CCL2 were upregulated in both young- and old-alcohol groups (Fig. S3E). Liver MOs isolated from the young-alcohol group exhibited a decrease in Tlr2 and Tlr8 signaling and an increase in calprotectin signaling (Fig. S3F). BM MOs from the old-alcohol group showed inhibition of antiviral responses, including Irf1, Irf7 and Irf9 (Fig. 3F). No pathways were retrieved for alcohol exposure in young BM MOs or any liver MØ populations due to the low number of DEGs in these groups (Fig. S3G and H).

We found no commonly dysregulated pathways across the four cell types with alcohol treatment in the young population. However, with aging, we observed that interferon (IFN) signaling was inhibited across microglia, liver MOs, and BM MOs (Fig. 3D-F).

Cell chat infers inter-organ MO/MØ communication and reveals unique crosstalk in response to aging and alcohol

Finally, we aimed to assess changes in the MO/MØ landscape across different organs using our transcriptomic data to infer inter-organ MO/MØ communications. For this, we used the Cellchat 2.1.2 package from R²⁰ and determined the strength of communication between the four cell types based on the ligand-receptor expression levels.

We observed an increased number of interactions (Fig. 4A) in the alcohol-treated groups (2,998 for young-alcohol and 2,916 for old-alcohol) compared to water-treated groups (2,733 for young-water and 1,648 for old-water), regardless of age. Self-signaling was stronger in liver MØs across all groups, which further increased upon alcohol treatment, while BM MOs showed the least number of interactions (Fig. 4A). Next, we analyzed the incoming (receiving) and outgoing (sending) signaling potential of the four cell types (Fig. 4B). Overall, liver MOs showed the highest incoming potential, while liver MØs showed the highest outgoing potential. BM MOs were the least communicative while microglia progressively increased their sending signals with age and alcohol (Fig. 4B). We observed that in the old mice, alcohol treatment had a striking effect on the outgoing or incoming strengths as opposed to the effect in young mice (Fig. 4B).

Fig. 4.

Cell chat infers inter-organ MO/MØ communication and reveals unique crosstalk in response to aging and alcohol. (A) Chord graph showing the total number of interactions among the four cell types in each experimental condition (young-water, young-alcohol, old-water, and old-alcohol). Tables below depict the exact number of interactions and quantify the interactions between two cell types from lowest (red) to highest (green). (B) Bubble plot illustrating outgoing (x axis) and incoming (y axis) signals for each cell type in each experimental condition. (C) Heat map from the top eight pathways triggering more outgoing cell-cell interactions in the experimental conditions. (D) Heat map from the top eight pathways triggering more inco1ming cell-cell interactions in the four experimental conditions. (E) Arrow graphs showing directional interactions between microglia, liver MOs, liver MØs, and BM MOs in ApoE, CCL, and SELPLG pathways. Transcriptomic analysis was performed in n = 4-6 mice per experimental condition and cell type. Color code for each cell type is microglia (blue), liver MOs (purple), liver MØs (red), and BM MOs (green). BM, bone marrow; MOs, monocytes; MØs, macrophages.

To further understand the nature of the signaling, we assessed the top 10 ligand-receptor interactions in each group (Table 1), identifying the top eight pathways that drive the most probable ligand-receptor communication as a sender (Fig. 4C) or recipient cell (Fig. 4D). APOE-TREM2 signaling from liver MØs to microglia was the top probable interaction across the four experimental groups. The maximum APOE production shifted from liver MØs to microglia in the old-alcohol group (Fig. 4C). With alcohol treatment, CC chemokine ligand (CCL) signaling was shifted from liver MOs to liver MØs in both young and old mice. This pathway also showed upregulation in microglia under old-alcohol conditions. Interestingly, FN1 (fibronectin 1) signaling was not present in the old-water group, highlighting its regulatory role in senescence.²¹ With age, protein S signaling, which is expressed by tumor-associated macrophages to induce MERTK- and TYRO3-induced inflammatory responses,²² intensified in liver MOs. For incoming signals, alcohol treatment in young and old mice led to increased CCL pathway signaling in liver MØs and BM MOs, while microglia showed the opposite trend. Complement pathway signaling increased in liver MØs in the young-alcohol group but decreased in liver and BM MOs in the old-alcohol group (Fig. 4D).

Table 1.

The top 10 ligand-receptor interaction pairs between the indicated MO/MØ cell types in the four treatment conditions.

	Source	Target	Ligand	Receptor	Probability	p-val	Pathway
Young-water
1	MØ liver	microglia	Apoe	TREM2_TYROBP	0.46	0	ApoE
2	microglia	MO liver	Selplg	Sell	0.33	0	SELPLG
3	microglia	MO liver	F11r	ITGAL_ITGB2	0.31	0	JAM
4	MO liver	microglia	Ccl5	Ccr5	0.31	0	CCL
5	microglia	MO BM	Selplg	Sell	0.31	0	SELPLG
6	microglia	microglia	Pros1	Mertk	0.29	0	PROS
7	MO BM	microglia	C3	ITGAM_ITGB2	0.29	0	COMPLEMENT
8	microglia	microglia	F11r	F11r	0.29	0	JAM
9	microglia	microglia	Apoe	TREM2_TYROBP	0.27	0.01	Apthe oE
10	MO BM	MØ liver	Fn1	Cd44	0.26	0	FN1
Young-alcohol
1	MØ liver	microglia	Apoe	TREM2_TYROBP	0.47	0	ApoE
2	microglia	microglia	Apoe	TREM2_TYROBP	0.40	0	ApoE
3	microglia	MO liver	Selplg	Sell	0.37	0	SELPLG
4	MO liver	microglia	Ccl5	Ccr5	0.35	0	CCL
5	MØ liver	MO liver	H2-D1	Cd8b1	0.32	0	MHC-1
6	microglia	MO BM	Selplg	Sell	0.32	0	SELPLG
7	MO BM	microglia	C3	ITGAM_ITGB2	0.32	0	COMPLEMENT
8	MO liver	MO liver	H2-D1	Cd8b1	0.31	0	MHC-1
9	MO liver	MO liver	H2-D1	Cd8b1	0.31	0	MHC-1
10	MO BM	microglia	C3	C3ar1	0.29	0	COMPLEMENT
Old-water
1	MØ liver	microglia	Apoe	TREM2_TYROBP	0.44	0	ApoE
2	microglia	MO liver	Selplg	Sell	0.35	0	SELPLG
3	MO liver	microglia	the Ccl5	Ccr5	0.34	0	CCL
4	microglia	MO liver	F11r	ITGAL_ITGB2	0.30	0	JAM
5	MO BM	MØ liver	Fn1	Cd44	0.29	0	FN1
6	microglia	MO BM	Selplg	Sell	0.29	0.01	SELPLG
7	microglia	microglia	F11r	F11r	0.28	0	JAM
8	MO BM	microglia	C3	ITGAM_ITGB2	0.28	0	COMPLEMENT
9	MØ liver	MO liver	H2-D1	Cd8b1	0.28	0	MHC-1
10	microglia	microglia	Pros1	Mertk	0.27	0	PROS
Old-alcohol
1	MØ liver	microglia	Apoe	TREM2_TYROBP	0.47	0	ApoE
2	microglia	microglia	Apoe	TREM2_TYROBP	0.46	0	ApoE
3	MO liver	microglia	Ccl5	Ccr5	0.39	0	CCL
4	MØ liver	MO liver	H2-D1	Cd8b1	0.38	0	MHC-1
5	MO liver	MO liver	H2-D1	Cd8b1	0.38	0	MHC-1
6	microglia	MO liver	Selplg	Sell	0.38	0	SELPLG
7	MO liver	MO liver	H2–K1	Cd8b1	0.37	0	MHC-1
8	MØ liver	MØ liver	Fn1	Cd44	0.33	0	FN1
9	MO BM	MO liver	H2-D1	Cd8b1	0.33	0	MHC-1
10	MO BM	microglia	C3	ITGAM_ITGB2	0.33	0	COMPLEMENT

Further pathways analysis showed that APOE signaling by BM MOs is diminished in young-alcohol and old-water groups compared to the young-water group. However, alcohol treatment in old mice induced new signaling from microglia to liver MØs (Fig. 4E top). Interestingly, we see the emergence of new communications between microglia and liver MØs along the APOE axis in the old-alcohol group. The CCL pathway showed a more complex autocrine signaling and interactions between the cell types (Fig. 4E middle). With alcohol treatment, we observed increased CCL ligand production by microglia in the young-alcohol group, which increased even further in the old-alcohol group. For example, we saw the emergence of a new signaling axis between microglia and liver MOs with alcohol treatment in the young-alcohol group; the strength of this interaction increased further in the old-alcohol group (Fig. 4E middle). Interestingly, in the old-alcohol group, we found diminished communication between BM MOs and liver MOs compared to the rest of the groups, as well as increased microglia signaling to all other MO/MØ populations. Similarly, the selectin P ligand (SELPLG) pathway showed increased communication from liver MØs to BM MOs in the old-alcohol group (Fig. 4E bottom).

Finally, we identified distinct ligand-receptor interactions associated with aging, alcohol exposure, and unique crosstalk specific to the young-water and old-alcohol groups. Table 2 summarizes the number of interactions for a given pathway in our four experimental conditions. Exclusive pathways found in old groups included ADIPONECTIN, BMP10, CSPG4, GDNF, HCRT, INSULIN, and MK, while 27HC, CDH5, CNTN, CX3C, GALANIN, NEGR, TAC, VISTA, and XCR were exclusively found in young groups, disappearing with age. In response to alcohol, we observed new interactions related to ACH, ADGRB, CD46, EPHA, NT, TAFA, and TRIIODOTHYRONINE pathways and the disappearance of NGF and CD23 interactions. Lastly, we identified exclusive pathways appearing in the old-alcohol group: AMH, IFN-III, IL17, and LEP.MSTN, NPVF, NRXN, OSM, PRLH, RBP4, SAA, SLITRK, SPP1, UROTENSIN, and VEGI pathways. CD200 and IFN-1 were lost exclusively in this same group.

Table 2.

The table enlists unique communication pathways and the number of interactions in the respective pathways, under different treatment conditions.

Pathway	Young water	Young alcohol	Old water	Old alcohol
Exclusive in young water
BRADYKININ	3	—	—	—
CCK	1	—	—	—
DMP1	3	—	—	—
ENHO	2	—	—	—
ESAM	3	—	—	—
GHRH	4	—	—	—
GPR	4	—	—	—
MELANOCORTIN	2	—	—	—
NPR2	2	—	—	—
SN	3	—	—	—
Not found in young water
Aldo	—	6	1	7
CD39	—	12	7	14
Cort	—	6	1	7
PSAP	—	4	4	4
SEMA3	—	63	13	18
SEMA5	—	10	9	5
TWEAK	—	4	2	4
Exclusive in old groups
ADIPONECTIN	—	—	3	8
BMP10	—	—	5	4
CSPG4	—	—	2	3
GDNF	—	—	3	2
HCRT	—	—	1	2
INSULIN	—	—	4	4
MK	—	—	10	22
Not found in old groups
27HC	2	2	—	—
CDH5	3	3	—	—
CNTN	7	14	—	—
CX3C	2	1	—	—
GALANIN	2	3	—	—
NEGR	3	3	—	—
TAC	2	4	—	—
VISTA	2	3	—	—
XCR	2	2	—	—
Exclusive in old alcohol
AMH	—	—	—	4
IFN-lII	—	—	—	2
IL17	—	—	—	3
LEP	—	—	—	3
MSTN	—	—	—	8
NPVF	—	—	—	1
NRXN	—	—	—	10
OSM	—	—	—	4
PRLH	—	—	—	1
PTH	—	—	—	2
RBP4	—	—	—	2
SAA	—	—	—	2
SLITRK	—	—	—	2
SPP1	—	—	—	25
UROTENSIN	—	—	—	1
VEGI		—	—	2
Not found in old alcohol
CD200	10	8	3	—
IFN-I	4	4	3	—
Exclusive in alcohol groups
Ach	—	6	—	7
ADGRB	—	18	—	4
CD46	—	2	—	1
EPHA	—	15	—	15
NT	—	7	—	10
TAFA	—	1	—	2
Triiodothyronine	—	6	—	3
Not found in alcohol groups
NGF	2	—	4	—
CD23	9	—	8	—

Histological evidence for APOE, TREM2, NRXN2, and SPP1 MOs/MØs crosstalk

In order to validate the effects of alcohol misuse and aging on intercellular communication, we performed immunofluorescence staining on liver and brain sections from various groups. We assessed expression of APOE (ligand) in the liver as it consistently ranked as one of the top communication pathways in our analysis (Fig. 4C,D, Table 1, Table 2) and changed its pattern of communication in the old-alcohol group (Fig. 4E). Since our transcriptomics data showed TREM2 in the microglia as the potential recipient of APOE signaling (Table 1, Table 2), we evaluated TREM2 levels in the brain sections. We identified microglia and liver MOs/MØs using the most commonly used markers, i.e. IBA-1 and CD68, respectively. As predicted by the transcriptomic analysis, we observed significantly higher APOE⁺ CD68⁺ MOs/MØs in the livers of the old-alcohol group compared to the old-water or young-alcohol group (Fig. 5A,E). Aging alone also had a significant effect on APOE expression in liver MO/MØ population (Fig. 5A,E). In the brain sections, microglia from the old-alcohol group showed significantly higher TREM2 expression compared to the young-alcohol groups (Fig. 5B,F). Aging alone also elevated TREM2 expression in microglia (as measured by two-way ANOVA for the effect of aging, p = 0.0017) (Fig. 5B,F).

Fig. 5.

Histological evidence for APOE, TREM2, NRXN2, and SPP1 MOs/MØs crosstalk. (A) Representative images of sections stained for APOE and CD68 maker for MOs/MØs in the liver. (B) Representative images of sections stained for TREM2, and IBA1 maker for microglia in the brain. (C) Representative images of sections stained for NRXN2, and CD68 maker for MOs/MØs in the liver. (D) Representative images of sections stained for SPP1, and IBA1 maker for microglia in the brain. (E) Quantification of CD68+ macrophages staining positive for APOE, determined per unit area (173 mm²) in images in A. (F) Quantification of mean intensity of TREM2 in IBA1+ microglia in B. (G) Quantification of CD68+ macrophages staining positive for NRXN2, determined per unit area (173 mm²) in images in C. (H) Quantification of mean intensity of SPP1 in IBA1+ microglia in D. (n ≥5 animals, 1 section per animal per group). Statistical significance was determined using two-way ANOVA; ∗p <0.05, ∗∗p <0.001, ∗∗∗p <0.0001, ∗∗∗∗p <0.00001, with Tukey’s post hoc comparison. MOs, monocytes; MØs, macrophages.

We also validated communication pathways specific to the old-alcohol group alone. To this end, based on the pathways with the maximum number of interactions, we focused on NRXN2 in liver MOs/MØs and SPP1 in microglia (Table 2). We observed significantly higher NRXN2⁺ CD68⁺ expression in MOs/MØs in the liver of the old-alcohol group compared to the old-water and young-alcohol groups (Fig. 5C,G). NRXN2 expression was also significantly higher in the old-water compared to the young-water group (Fig. 5C,G). SPP1 showed a general effect of alcohol (as evidenced from the two-way ANOVA for the effect of treatment, p = 0.0016) (Fig. 5D,H). We observed a trend towards increased SPP1 with alcohol in both young-alcohol and old-alcohol groups compared to respective water controls (Fig. 5D,H).

Taken together, we found a strong shift in autocrine and paracrine signaling with aging and alcohol treatment that showed influences on transcriptomic markers of MO/MØ inter-organ communication pathways.

Alcohol misuse in aging increased circulating cytokines and reduced phagocytosis signatures

Inflammaging is a hallmark of age-associated immune dysfunction, marked by heightened cytokine response and reduced phagocytic potential of MØs (4). We also analyzed expression of serum cytokines in all groups and observed that IL1β showed a significant increase in the old-alcohol group compared to the old-water group, while GM-CSF showed a strong increasing trend in the old-alcohol compared to the young-alcohol groups (Fig. 6A, E). Other cytokines (IL6, MCP1, IL2, TNFα) showed increased production due to the effect of aging in general, consistent with the notion of inflammaging (Fig. 6B-D,F) (4). To assess phagocytic potential, we turned towards transcriptomic data and analyzed the expression of genes associated with phagocytosis (gene list from Gene Ontology term analysis). We conducted GSEA to determine whether the set of phagocytosis-relevant genes (selected based on Gene Ontology pathways list) was significantly modulated in our populations. In the young groups, we did not observe a significant effect of alcohol on phagocytosis (6G, bottom panel); however, we observed significant alcohol-induced downregulation of phagocytosis in most cell types in the old-alcohol group compared to the old-water group (Fig. 6G, top panel). This comparison showed a mere trend towards reduced phagocytosis in microglia due to alcohol in aging conditions and did not reach significance. Overall, these data together suggest that alcohol misuse worsened age-associated immune changes with respect to cytokine production and MO/MØ phagocytic potential (Fig. 6G).

Fig. 6.

Alcohol misuse in aging increased circulating cytokines and reduced phagocytosis signatures. (A-F) Quantification of cytokines measured in the serum of mice from various groups. (n ≥4 animals, per group). Statistical significance was determined using two-way ANOVA; ∗ p <0.05, ∗∗p <0.001, ∗∗∗p <0.0001, ∗∗∗∗p <0.00001, with Sidak post hoc comparison. (G) (Top panel), GSEA of the phagocytosis gene list (GO pathway) comparing results from the indicated cell type in old-alcohol vs. old-water groups; (Bottom panel) GSEA of the phagocytosis gene list (GO pathway) comparing results from the indicated cell type in young-alcohol vs. young-water groups. BM, bone marrow; GO, Gene Ontology; GSEA, gene set-enrichment analysis; MOs, monocytes; MØs, macrophages.

Discussion

MOs/MØs are plastic and heterogeneous immune cells found in most organs, where they perform homeostatic, surveillance, and defense functions. As key contributors in innate immune responses, MOs/MØs are known to respond to pathogen- and damage-derived inflammatory signals, promote organ tolerance, and induce sterile inflammation. In patients with alcohol misuse, both age and alcohol consumption independently shift monocyte phenotype towards heightened inflammatory states with compromised antimicrobial defenses.²³ Alcohol misuse has been shown to affect all organs, particularly the liver, brain, and the immune system.^16,[24], [25], [26], [27] Alcohol misuse affects Kupffer cells (liver MØs) and leads to recruitment of inflammatory MOs to the liver,²⁸ as well as activating microglia in the brain.¹⁶ While KCs and microglia are yolk-sack derived, the BM replenishes circulating MOs that are recruited to sites of injury. However, the effects of aging and alcohol on the regulation of these MOs/MØs are unknown. Here, we provide the first glimpse of the direct comparison between the MO/MØ landscape across three organs during aging and alcohol misuse. Importantly, we show that alcohol metabolism is impaired in aged mice mainly due to downregulation of CYP2E1 levels. Specifically, blood alcohol and its toxic metabolites, including acetaldehyde and acetate, remain elevated in the circulation for a longer time than in young mice, further contributing to inflammaging in the aged mice.

We identified a MO/MØ aging signature of seven common DEGs, emphasizing the stronger effect of the cellular niche than the cell type origin. Amongst these, Cd22²⁹ and Ide³⁰ have previously been associated with old microglia and their phagocytic capabilities. The four downregulated genes (Dynlt1b, Wdfy1, Fkbp5, and Serinc3) have been implicated in detection/response to viruses and bacteria and regulation of toll-like receptor 3/4, highlighting a likely impaired defense response of these cells against pathogens and cancer.[31], [32], [33], [34]

Despite the marginal overlap in DEGs, old MO/MØ populations showed a prominent activation of viral response and pathways leading to inflammation (NOD-like receptor signaling, JAK-STAT signaling, and TNF-NFκB signaling), reminiscent of inflammaging. The relaxin pathway, which appeared uniquely downregulated in old microglia, has been associated with anti-fibrotic effects in the liver³⁵ and anti-inflammatory effects on microglia and macrophages in models of stroke and hemorrhage.[36], [37], [38] Thus, reduced relaxin signaling may contribute to the pro-inflammatory state of microglia in aging.

Interestingly, alcohol misuse had no common DEGs across the four MO/MØ populations, and IFN inhibition was a common pathway in MOs/MØs from the old-alcohol group. IFN signaling is critical in regulating adaptive immunity during viral infections,^39,40 partially explaining the increased susceptibility to viral infections (e.g. COVID-19 and influenza) in the aged population with alcohol misuse. Older individuals with alcohol misuse are more susceptible to complications and increased incidence of viral infections, such as COVID-19 and influenza.^11,41

Liver MØs showed no response to alcohol, either in the young or old groups. On the other hand, BM MOs were non-responsive to alcohol treatment in the young, but were the most alcohol-responsive in the old cohort. This dramatic shift may indicate age-related susceptibility to alcohol and/or be a result of higher BAC affecting the BM in old mice.

Since our assay involves three different alcohol-relevant microenvironments, we determined the probability of communication between the different MO/MØ populations based on the expression of ligand/receptor pairs. We summarized the 10 interactions into 8 pathways, including APOE, SELPLG, CCL, JAM, FN1, COMPLEMENT, MHC-1, and protein S. Of note, SPP1 (osteopontin) pathway components were selectively expressed in the old-alcohol condition, while those of CD200 and IFN-1 disappeared in the old-alcohol group. Understanding inter-organ communication in models of aging is critical, particularly given the fact that immune aging has a direct impact on the biological age of other body systems.⁵ Moreover, inter-organ communication is of high relevance during alcohol-induced inflammation. Our work has previously shown that alcohol-induced systemic inflammation⁴² affects different organs through circulating cytokines, MOs, and extracellular vesicles.⁴³

Lastly, we observed that APOE signaling was one of the top pathways that increased in signaling strength with age. The homozygous APOE4 haplotype exhibits near-complete penetrance for Alzheimer’s pathology and confers a 13-fold increased risk of developing Alzheimer’s disease.^44,45 In addition, APOE is a major risk factor for age-associated atherosclerosis and cardiovascular disorders.⁴⁶ Interestingly, our data shows that alcohol misuse accelerates the aging phenotype through pathways such as APOE, specifically from the crosstalk between liver-MØs > microglia or microglia > microglia, opening new avenues for exploring APOE as a driver for liver-brain crosstalk in the context of alcohol-related liver disease.

We acknowledge certain limitations in this study. First, we realize that the alcohol-induced MO/MØ responses may be limited by the use of the C57BL6/N genetic background, choice of the alcohol paradigm, and the use of female mice only. Genetic background and different alcohol regimens have been shown to differentially induce organ damage.⁴⁷ Regarding sex, females are more susceptible to alcohol-induced liver damage.⁴⁸ Second, human studies examining monocyte populations with well-documented data on aging and alcohol misuse are lacking, which limits our ability to validate these findings clinically.

In summary, this manuscript describes the MO/MØ transcriptomic landscape in a model of alcohol misuse and aging, and identifies novel and unique communication pathways between MO/MØ populations in different organ niches, paving the way for better understanding of the inter-organ aging network in health and in the context of alcohol misuse.

Abbreviations

BAC, blood alcohol concentration; BM, bone marrow; CCL, CC chemokine ligand; DEG, differentially expressed gene; GSEA, gene set-enrichment analysis; IFN, interferon; IL, interleukin; KCs, Kupffer cells; MOs, monocytes; MØ, macrophages.

Financial support

This study was supported by NIH grants R01 AA011576, R01 AA017729 and R01 AG072899 (to GS).

Authors’ contributions

MOR, RJ and GS designed the experiments, analyzed and interpreted the data, and wrote the manuscript. SGM analyzed and discussed the data. VB and PTN conducted experiments, analyzed and discussed the data. VB, YZ, AP and MB conducted the experiments. JGS discussed data. GS obtained funding. All authors critically reviewed the manuscript.

Data availability

The data supporting the findings of this study are available within the article and its supplementary materials. RNA-seq data that support the findings of this study will be deposited in Gene Expression Omnibus (GEO) and made publicly available upon acceptance of the manuscript.

Conflict of interest

GS reports being a paid consultant for Evive Bio, Merck, Durect Corporation, Terra Firma, Pandion Therapeutics, Satellite Bio, Cyta Therapeutics, Pfizer, Intercept, Surrozen, Resolution, Boehringer Ingelheim and NovoNordisk. She holds stock options in Glympse and Zomagen Bioscience/Ventyx Biosciences and receives royalties from Springer Nature Group and UpToDate Inc. The other authors declare no conflicts of interest to the research conducted.

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

Appendix A. Supplementary data

The following are the Supplementary data to this article.