Article Text
Abstract
Objective Bone marrow-derived myeloid cells accumulate in the liver as monocytes and macrophages during the progression of obesity-related non-alcoholic fatty liver disease (NAFLD) to steatohepatitis (NASH). Myeloid cells comprise heterogeneous subsets, and dietary overnutrition may affect macrophages in the liver and bone marrow. We therefore aimed at characterising in depth the functional adaptations of myeloid cells in fatty liver.
Design We employed single-cell RNA sequencing to comprehensively assess the heterogeneity of myeloid cells in the liver and bone marrow during NAFLD, by analysing C57BL/6 mice fed with a high-fat, high-sugar, high-cholesterol ‘Western diet’ for 16 weeks. We also characterised NAFLD-driven functional adaptations of macrophages in vitro and their functional relevance during steatohepatitis in vivo.
Results Single-cell RNA sequencing identified distinct myeloid cell clusters in the liver and bone marrow. In both compartments, monocyte-derived populations were largely expanded in NASH-affected mice. Importantly, the liver myeloid compartment adapted a unique inflammatory phenotype during NAFLD progression, exemplarily characterised by downregulated inflammatory calprotectin (S100A8/A9) in macrophage and dendritic cell subsets. This distinctive gene signature was also found in their bone marrow precursors. The NASH myeloid phenotype was principally recapitulated by in vitro exposure of bone marrow-derived macrophages with fatty acids, depended on toll-like receptor 4 signalling and defined a characteristic response pattern to lipopolysaccharide stimulation. This imprinted and stable NASH myeloid immune phenotype functionally determined inflammatory responses following acute liver injury (acetaminophen poisoning) in vivo.
Conclusion Liver myeloid leucocytes and their bone marrow precursors adapt a common and functionally relevant inflammatory signature during NAFLD progression.
- liver
- macrophages
- monocytes
- NAFLD
- fatty acids
- inflammation
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Infiltration of bone marrow monocytes into the liver drives the progression of hepatic steatosis towards steatohepatitis and fibrosis in obesity.
Recent publications have highlighted the concept of functional adaptations on metabolic stimuli in myeloid immune cells during non-alcoholic fatty liver disease (NAFLD).
What are the new findings?
Single-cell RNA sequencing identifies hallmark changes of a common ‘NAFLD myeloid phenotype’, including the downregulated gene expression of the inflammatory marker calprotectin (S100a8/S100a9) on Western diet feeding in the liver monocytes, macrophages and dendritic cells, as well as in their bone marrow precursors.
Western diet feeding induces a distinct gene signature common to all myeloid leucocyte subsets that can be largely replicated in vitro by prestimulation with free fatty acids of bone marrow-derived macrophages in a toll-like receptor 4-dependent manner.
The ‘NAFLD myeloid phenotype’ is characterised by changes in the inflammatory capacity of bone marrow monocytes, which remain stable, even on changes to their local micromilieu or after in vitro cytokine stimulation, and can be adoptively transferred.
How might it impact on clinical practice in the foreseeable future?
Our data provide a more detailed insight into the complex adaptation mechanisms of liver macrophages during the development of obesity-related steatohepatitis, which might support improved risk stratification of patients at different stages of the disease and facilitate the development of novel therapeutic strategies.
Introduction
Non-alcoholic fatty liver disease (NAFLD) due to obesity and metabolic disorders is the most prevalent liver disease worldwide.1 During NAFLD progression, the capacity of the liver to metabolise energy substrates such as carbohydrates and fatty acids (FAs) is overwhelmed, leading to accumulation of toxic lipid metabolites, hepatocytic lipid storage, hepatocellular stress, cell death, and subsequently activation of immune mechanisms.2 Immune cell infiltration into the liver is one hallmark characteristic of non-alcoholic steatohepatitis (NASH), the progressive inflammatory form of NAFLD. Bone marrow-derived monocyte influx into the liver has been associated with ongoing hepatic inflammation and NAFLD progression.3 4 Monocytes and monocyte-derived macrophages (MoMF) are known for their high plasticity, allowing them to adapt their inflammatory polarisation dependent on the simultaneous recognition of signals like cytokines, chemokines and damage-associated molecular patterns (DAMPs).5 6 Prior studies have highlighted that FAs act as inflammatory mediators, shaping a unique inflammatory phenotype in human MoMF in vitro.7 The composition of circulating FA shows a high variation depending on the individual dietary intake. Recognition of FA by mononuclear cells is thought to depend on toll-like receptor 4 (TLR4) and nuclear factor kappa B signalling, followed by subsequent NOD-like receptor family, pyrin domain-containing 3 (NLRP3) inflammasome formation and activation.8 In line, TLR4-mediated lipopolysaccharide (LPS) recognition has been shown to induce a ‘myeloid over lymphoid’ bias in bone marrow precursors,9 and dietary changes during Western diet (WD) feeding alter the response towards LPS even after switching to normal standard diet.10 While these studies established the principal link between metabolic disease progression and inflammatory macrophage polarisation, the link between inflammatory polarisation of bone marrow precursors and the subsequent influence on liver monocytes, macrophages and dendritic cells (DCs) during NAFLD progression is less clear.
In our study, we demonstrate that bone marrow monocytes (BMM) and precursors acquire characteristic changes in their gene expression pattern during diet-induced NAFLD progression, which are preserved in and determine the function of hepatic myeloid cell populations. Most prominently, calprotectin coding genes S100 calcium-binding protein A8 (S100a8) and S100a9 are downregulated in myeloid immune cells in both the liver and bone marrow. The imprinted changes in myeloid immune cells affected their inflammatory polarisation and modulated the response to acute sterile liver injury. Thus, our data suggest that NAFLD progression induces unique, stable and functionally relevant changes of the inflammatory immune response in myeloid cells in the bone marrow and liver.
Materials and methods
Animal experiments
C57BL6/J wildtype or Ccr2−/ − (research resource identifier (RRID): IMSR_JAX:017586) mice were housed in a specific pathogen-free environment. Fatty liver injury was induced by feeding WD with high levels of sugar (34%), fat (21%) and cholesterol (1.25%) (ssniff, Soest, Germany; product no E15723-34) for 16 weeks. Control mice received low-fat normal diet (ND). As a standard model of acute liver injury, ND-fed or WD-fed animals were subjected to an overdose of acetaminophen (APAP) (250 mg per kg body weight intravenously, after 12 hours of fasting).11 All in vivo experiments were performed with male mice at 8–24 weeks of age.
Phenotypic assessment
Conventional H&E or Sirius Red staining was performed on paraffin-embedded liver and adipose tissue (AT) sections. Hepatic leucocytes were isolated by collagenase digestion and differential centrifugation followed by multicolour flow cytometry using an LSRFortessa (BD Biosciences), as previously described.11 Gut microbiome composition was assessed by 16S rRNA sequencing of colon samples at the ZIEL - Institute for Food & Health (Technical University Munich, Germany) and analysed by integrated microbial next generation sequencing (IMNGS) and Rhea.12
Single-cell RNA sequencing analysis of hepatic and bone marrow myeloid cells
Myeloid leucocytes were isolated from the liver and bone marrow of ND-fed and WD-fed mice and sorted as CD45+ LY6G− TCRβ− CD4− CD19− NK1.1− by fluorescence activated cell sorting (FACS) using Aria-II (BD Biosciences). Isolated cells were washed once with cold phosphate-buffered saline (PBS) + 0.1% bovine serum albumin (BSA) and subjected to single-cell RNA sequencing analysis using the Chromium System (10x Genomics, California, USA). Per sample, 5000 cells were used with a recovery rate of about 60%. Sequencing was performed on Illumina NextSeq 550 (paired-ends, 2×75 bp), resulting in 60 000 reads per cell and ~4×108 total reads per Illumina chip. Primary data analysis was performed with an inhouse pipeline embedded in the workflow management system of the QuickNGS environment,13 based on 10x Genomics Cell Ranger. Final analysis of the single-cell gene expression data was done with the Seurat package for R.14
Bone marrow-derived macrophage polarisation analysis and bulk RNA sequencing
To generate bone marrow-derived macrophages (BMDM), bone marrow from ND-fed or WD-fed mice was flushed with cold RPMI supplemented with 10% fetal calf serum (FCS), 1% Penicillin/Streptomycin and 30% L929 supernatant. Afterwards, cells were incubated for 24 hours at 37°C. For FA prestimulation the medium was supplemented with 0.1 mM mix of palmitic and oleic acid (1:3) dissolved in FA-free BSA as previously described.3 For TLR4 inhibition, 200 nM of the small molecular inhibitor TAK-242 (Sigma, Germany) dissolved in dimethyl sulfoxide (DMSO) was added. Final concentration of DMSO was below 0.1% in all conditions. After 4 days, cells were washed and seeded at a density of 5×105 cells/well on a 6-well plate and incubated for another 3 days at 37°C. On day 7, BMDM were stimulated with 100 ng/mL of Escherichia coli-derived LPS for 24 hours. Cells were collected for flow cytometry and RNA isolation. BMDM were lysed in peqGold (VWR, Germany), and total RNA was isolated by phenol-chloroform extraction. Gene expression analysis was done by mRNA sequencing on the Illumina NextSeq 550 (TruSeq Stranded mRNA Kit, 1×125 bp). Cytokines were measured from supernatant by the LUNARIS cytokine multiplex assay (Ayoxxa, Germany).
Statistics
All experimental data from mice are presented as mean±SD. Differences between groups were evaluated by unpaired one-way analysis of variance with Bonferroni multiple comparisons correction (GraphPad Prism, GraphPad Software, USA).
Additional information is provided in online supplementary methods.
Supplemental material
Results
Single-cell RNA sequencing reveals the profound expansion of distinct liver infiltrating myeloid leucocyte subsets during obesity-related fatty liver disease
The recruitment of CCR2+ bone marrow-derived monocytes into the liver (and AT) has been identified as a hallmark characteristic of NAFLD progression due to metabolic syndrome in mouse models4 15 16 and human disease.3 17 We herein employed the model of feeding mice a high-carbohydrate, high-fat, high-cholesterol ‘Western diet’ or normal diet for 16 weeks, which resulted in typical macroscopic steatosis, histological steatohepatitis and fibrosis with prominent perisinusoidal and periportal macrophage infiltrates (figure 1A,B). WD feeding induced features of NASH, as corroborated by increased serum alanine aminotransferase (ALT), aspartate aminotransferase and glutamate dehydrogenase (GLDH) levels (figure 1C), liver to body weight ratio, and liver triglycerides (figure 1D). The populations of MoMF in the liver, defined as hepatic CD11b+F4/80+ cells by flow cytometry, as well as monocytes in bone marrow, progressively increased in WD-fed mice (figure 1E). Importantly, feeding a Western-style diet to mice induced profound changes characteristic of metabolic syndrome in extrahepatic tissues as well. Along changes in the liver, WD also induced swelling of the adipocytes in the white AT, as well as an increase in the absolute and relative weight of perigonadal AT per body weight (figure 1F). Another hallmark feature of metabolic syndrome are alterations in the gut microbiome, with a relative reduction of Bacteroidetes and an increase of Actinobacteria and Verrucomicrobia (figure 1G).
For a thorough and comprehensive analysis of adaptations in myeloid cell populations, we conducted single-cell RNA sequencing of FACS isolated LY6Gneg CD45pos non-lymphoid leucocytes from both liver and bone marrow. Lymphoid populations were excluded by staining for T cells, B cells, natural killer (NK) and NKT cells (CD3, CD4, CD19, NK1.1), and neutrophils were excluded based on LY6G expression (figure 2A). Single-cell RNA sequencing data were then projected onto t-SNE plots, where cells are clustered together that resemble each other in their overall gene expression. Distinct myeloid cell clusters were then annotated by uniquely expressed markers (figure 2B,C; see online supplementary table 1). Hepatic MoMF and BMM were identified by a high expression of lysozyme 2 (Lyz2), while Kupffer cells (KC) were annotated by a high C-type lectin domain family 4, member F (Clec4f) expression. DC subsets were split into chemokine (X-C motif) receptor 1 (Xcr1) expressing conventional type 1 DC (cDC1), Cd209a or Ccr7 expressing conventional type 2 DC (cDC2) and plasmacytoid DC highly expressing sialic acid binding Ig-like lectin H (Siglech) (figure 2B). We found the KC population to be under-represented in our single-cell RNA sequencing data set, as compared with data obtained by flow cytometry, possibly reflecting a reduced yield due to the isolation procedure. In the bone marrow, PCNA clamp associated factor (Pclaf) positive-common monocyte progenitors (CMoP) and Cd34-positive haematopoietic stem cells (HSC) were identified. In addition, bone marrow DC precursors were annotated by a high expression of Cd209a, interferon regulatory factor 8 (Irf8) and Siglech (figure 2C). Bone marrow populations displayed a reduced heterogeneity between ND and WD as compared with their liver counterparts (figure 2B,C). Thus, myeloid cells of bone marrow origin profoundly accumulate in the liver during development of obesity-related NAFLD, and single-cell RNA sequencing analysis identified previously unrecognised diverse myeloid cell clusters in both compartments.
Liver macrophage subsets and BMM share characteristic gene expression patterns in fatty liver disease
The striking heterogeneity of myeloid cells in the liver and bone marrow prompted us to better characterise these subclusters and their potential role during WD-induced obesity, particularly the monocyte-derived subsets in the liver and bone marrow. The cluster of MoMF I showed a high expression of the extracellular matrix protein fibronectin 1 (Fn1), as well as microsomal glutathione S-transferase 1 (Mgst1) and methionine sulfoxide reductase B1 (Msrb1), both associated with oxidative stress. MoMF II expressed few marker genes including Chil1, mainly indicating a bone marrow origin. MoMF III showed an inflammatory activated state, indicated by a high expression of Il1b (figure 3A). The relative numbers of MoMF I were almost identical between ND and WD, whereas MoMF II and III were found to be over-represented in the livers after WD (figure 3B). Next, we analysed the top regulated genes for each cluster. We found that S100a8 and S100a9, the two heterodimeric subunits of the inflammatory marker calprotectin, as well as inflammation-associated resistin like gamma (Retnlg) and secretory leucocyte peptidase inhibitor (Slpi), were all highly expressed in MoMF isolated from livers of ND mice compared with WD mice (figure 3C). Furthermore, perilipin 2 (Plin2), associated with lipid droplet formation and triglyceride metabolism, was uniquely upregulated in MoMF I and III on WD feeding (figure 3C). Confirmatory to the downregulated gene expression of calprotectin, we validated these findings by a reduced protein expression of S100A9 in liver MoMF by FACS (figure 3D and see online supplementary figure 1A–D).
Compared with liver MoMF, BMM populations showed fewer differentially regulated marker genes. BMM I closely resembled cluster MoMF I in the liver by high expression of Mgst1. Both BMM I and BMM II expressed C-type lectin domain family 4, member a3 (Clec4a3), underlining their monocyte phenotype.18 The second cluster of BMM was additionally characterised by high expression of G protein subunit gamma transducing activity polypeptide 2E (Gngt2) and signal-regulatory protein beta 1C (Sirpb1c). Cluster BMM III was found to show high expression of centromere protein A (Cenpa) and cyclin B2 (Ccnb2), both associated with cell cycle regulation (figure 3E). Between the dietary conditions, the three BMM populations showed less pronounced changes in relative abundance than liver MoMF (figure 3F). However, S100a8 and S100a9 were again significantly higher expressed in ND compared with WD conditions in BMM I and S100a9 in BMM III (figure 3G,H). Taken together, these data indicate a common phenotype of differential gene expression in both liver MoMF and BMM, characterised, among others, by a strongly reduced expression of inflammatory calprotectin.
Western-style diet feeding induces characteristic changes in gene expression profiles among myeloid leucocytes across all cellular subtypes
Analogous to the RNA expression pattern in MoMF and BMM during NAFLD progression, we analysed significantly regulated genes for all myeloid cell clusters in the liver and bone marrow, including KC, cDC2 (both in liver), as well as CMoP and HSC (both in bone marrow), between ND and WD feeding conditions. S100a8 and S100a9 were again significantly reduced on WD feeding in liver KC and cDC2 (figure 4A), as well as in CMoP but not in HSC (figure 4B). Since other genes (besides calprotectin) were also similarly regulated in different subpopulations between feeding conditions, we next analysed the OR of shared genes between every myeloid cell cluster. Remarkably, we found that every myeloid cell cluster in the liver (figure 4C) and bone marrow (figure 4D) shared significantly regulated genes dependent on the dietary condition (ie, WD vs ND feeding), irrespective of the cellular subtype.
For the overall regulation among all clusters, bone marrow leucocytes showed fewer significantly regulated genes between the WD and ND conditions compared with liver, while a considerable number of genes were coregulated between both compartments in either ND or WD condition (figure 4E). The analysis of shared significantly over-represented gene ontology (GO) pathways revealed that inflammatory migration of leucocytes was strongly associated with ND feeding in both liver and bone marrow, whereas metabolic GO terms such as cellular amide metabolic or nitrogen biosynthetic processes were upregulated in the majority of all myeloid leucocyte populations in the liver and bone marrow on WD feeding. While genes and GO terms were shared between liver and bone marrow leucocytes within ND or WD, there was no overlap of significantly enriched genes or GO terms in between all diets and clusters (figure 4F). These data demonstrate the imprinting of a common ‘NAFLD gene signature’ in myeloid leucocytes in both liver and bone marrow, with more pronounced changes of myeloid liver leucocyte phenotypes compared with their bone marrow counterparts.
Differences in the inflammatory capacity of myeloid cells are already induced in the bone marrow
The detection of the ‘NAFLD gene signature’ in myeloid bone marrow cells indicated that the inflammatory capacity of monocytes in obese mice is already altered in the bone marrow. We therefore generated BMDM in vitro from either ND-fed or WD-fed mice. Purity and macrophage trait after 7 days of cultivation were confirmed by high expression of CD11b and F4/80, whereas LY6C was downregulated (figure 5A). After 7 days, BMDM were then stimulated with LPS for 24 hours followed by mRNA sequencing analysis. As seen in the multidimensional scaling plot, groups cluster depending on origin, either ND or WD, as well as LPS treatment, underlining the presence of stable and functionally relevant changes induced in BMM in obese mice (figure 5B). We next investigated whether classical LPS responses were altered between the two conditions. Interestingly we found a significantly increased secretion of interleukin 6 (IL6) between lean and obese mice, while gene expression levels of Il6 and Il1b remained unaltered following LPS stimulation (figure 5C). Most interestingly, BMDM from ND-fed mice showed a higher expression of the neutrophil recruiting chemokines Cxcl1, Cxcl2 and Cxcl3 following LPS stimulation, as compared with BMDM from WD mice, which showed a higher expression of Cxcl10 and Cxcl11 relevant for T cell recruitment. Moreover, Slpi as well as S100a8 were found to be significantly upregulated in BMDM from ND mice as compared with WD counterparts (figure 5C,D), agreeing with our in vivo data obtained by single-cell RNA sequencing analyses. Altogether, these data indicated that BMDM retain a unique and functionally relevant inflammatory activation profile, depending on whether they had been generated from ND-fed or WD-fed mice.
Imprinting of the common inflammatory immune phenotype in BMDM is mediated by FAs via TLR4 signalling
Signalling via TLR4 in macrophages has been proposed to link metabolic diseases with inflammation.19 We hypothesised that FAs are a main driver of the ‘metabolic phenotype’ in macrophages and tested the relevance of TLR4 for this mechanism by using the TLR4 inhibitor TAK-242 during FA pretreatment of BMDM for 7 days (figure 6A). RNA sequencing analysis showed that FA pretreatment (of BMM from ND-fed mice) indeed induced a gene expression pattern closely mimicking that of WD BMDM (figure 6B). TAK-242 exposure to FA-pretreated BMDM largely abrogated the specific response pattern towards LPS that was seen in BMDM pretreated with FA in comparison with untreated BMDM (figure 6B). Interestingly, we found that L-selectin (CD62L) was the strongest marker showing the differential regulation of FA pretreatment via TLR4 on both gene as well as protein expression level (figure 6C). While tumour necrosis factor alpha (TNF-α) secretion was reduced in both WD and FA-treated BMDM, IL6 secretion was found to be only significantly reduced in FA-pretreated BMDM (figure 6D). Both TNF-α and IL6 secretion depended on TLR4 signalling (figure 6D). Additionally, costimulation with oxidised LDL and FA was found to have a moderately additive effect on the increased expression of CD62L and a suppressive effect on the expression of the inflammatory marker CD86. For the secretion of IL6, TNF-α or IL10, we could, however, not observe a synergistic effect of oxidised LDL and FA (online supplementary figure 2A–C). Additional experiments using BMDM from Tlr4−/− mice confirmed the TLR4 dependency of the imprinting during FA prestimulation (online supplementary figure 2D). However, it needs to be noted that the LPS stimulation is also strongly attenuated in the absence of functional TLR4 signalling.
Differential gene expression analysis showed that neutrophil recruiting chemokines and Slpi were significantly upregulated with TLR4 inhibition during FA pretreatment following LPS stimulation, in agreement with our data from BMDM derived from ND-fed and WD-fed mice, while the expression of S100a8 or S100a9 did not depend on TLR4 signalling (figure 6E). The subsequent gene pathway analysis corroborated that the altered inflammatory response towards LPS following FA pretreatment was in part mediated by TLR4-dependent signalling, involving genes associated with acute inflammatory responses, neutrophil recruitment, lipid and ion metabolism, as well as modulation of extracellular matrix by matrix metalloproteinases (Mmp) (figure 6F). Thus, characteristic changes related to WD feeding in vivo could be principally mimicked by exposure of BMDM to FA ex vivo, suggesting that dietary FAs contribute to the NAFLD monocyte/macrophage phenotype. These data also indicated that TLR4-dependent macrophage reprogramming in response to FA is characterised by a reduced expression of chemokines for neutrophil recruitment as well as metabolic and matrix-modulating functions.
The common NAFLD monocyte phenotype determines the inflammatory response to acute liver injury in vivo
Acute liver injury results in the chemokine receptor CCR2-dependent recruitment of MoMF with characteristic inflammatory properties,20 which aggravate the early course (ie, first 12 hours) of experimentally induced liver damage following APAP poisoning in mice.11 We hypothesised that the unique inflammatory polarisation in myeloid cells induced by NAFLD progression would alter the response to injury signals and tested this hypothesis by challenging ND-fed versus WD-fed mice with a non-lethal overdose of APAP. We found that the acute liver injury following APAP overdose, as assessed by H&E staining and serum ALT and GLDH activity, was significantly attenuated in WD-fed mice as compared with equally treated ND mice (figure 7A,B). Following APAP overdose, bone marrow-derived monocytes (ie, MoMF) and neutrophils infiltrated into the liver in ND-fed mice, which was reduced in WD-fed mice (figure 7C). Importantly, the attenuated injury to APAP was related to inflammatory cells and not to alterations of drug metabolism, as confirmed by analysing key enzymes of APAP metabolism following WD feeding (online supplementary figure 2A,B). In line, hepatic S100a8 and S100a9 expression was found to be significantly reduced in WD-fed mice as compared with their ND-fed counterparts following APAP overdose (online supplementary figure 3C). Furthermore, we confirmed the principally similar APAP metabolism and liver injury induction between ND-fed and WD-fed mice by assessing that liver injury at 3 hours after APAP was comparable between ND and WD (online supplementary supplementary figure 3D), supporting an injury-modulating role of myeloid cells in WD-fed mice.
Based on this finding, we hypothesised that the NAFLD-related phenotypic adaptations in myeloid cells in liver and bone marrow prevented excessive inflammatory responses to acute liver injury. In order to prove that the BMM polarisation contributed to determining the course of acute liver injury, we intravenously injected 3×106 BMM from either ND or WD mice into either ND-fed or WD-fed mice, 2 hours after APAP administration (figure 7D and online supplementary figure 3E). The adoptive transfer of BMM from ND into WD APAP-challenged mice restored the population of S100A9+ MoMF, without affecting S100A9+ neutrophils (figure 7E). The adoptive transfer of BMM isolated from ND mice, but not WD mice, indeed led to an increased liver injury in WD-fed mice (figure 7F). The adoptive transfer of WD-BMM into ND-fed mice, on the contrary, did not reduce APAP injury (figure 7F). While the adoptive transfer of ND BMM into WD-fed mice restored S100A9+ MoMF, we could not observe an effect on the total numbers of MoMF, neutrophils, CD4 or CD8 T cells, and total hepatic levels of Cxcl1, Cxcl2 or Cxcl10 remained unaltered in the liver following the adoptive transfer of ND or WD BMM (online supplementary figure 3F,G). This indicates a direct effect of BMM during the reduced innate immune response following diet-induced obesity and steatohepatitis. Altogether, these data indicate a common and stable ‘metabolic phenotype’ in the liver and bone marrow myeloid cells, particularly monocytes, which determines the functional response to danger signals and prevent excessive inflammatory responses to hepatic injury.
Discussion
Chronic dietary overnutrition and obesity are common health problems in industrialised countries and regularly lead to obesity-associated diseases such as metabolic syndrome and NAFLD.1 Liver infiltrating monocytes are highly plastic regulators of the immune response and play a key role in the progression of NAFLD towards NASH and fibrosis.21 A recent study could also confirm the high heterogeneity and plasticity of macrophage populations in human liver.22 Free FAs have been shown to induce a unique inflammatory phenotype in both human7 and mouse monocytes.23 In our study, we characterised the unique inflammatory polarisation of myeloid cell subsets in the liver of WD-fed mice, which were mirrored by similar changes in the bone marrow compartment. Furthermore, the characteristic profile of BMDM from WD-fed mice could be partially recapitulated when BMDM were pretreated with FA, suggesting that nutritional saturated FAs are a major driver of the common NAFLD monocyte phenotype.24 Among the phenotypic changes in myeloid subsets during the progression of experimental NAFLD, we identified a set of genes consisting of S100a8 and S100a9 that were significantly regulated in the liver and bone marrow myeloid leucocytes.
Calprotectin functions as a DAMP and serves as a clinical marker in various inflammation-driven diseases such as rheumatoid arthritis or IBD.25 Polyunsaturated FAs were described to reduce the expression of calgranulin genes post-LPS stimulation in THP-1 monocytes in vitro.26 Earlier studies also described calprotectin as an FA carrier,27 which might explain its regulation during NAFLD progression. Calprotectin can be recognised via the TLR4–MyD88 axis,28 and the receptor for advanced glycosylation end products, thereby inducing NLRP3 inflammasome formation.29 Of note, we found that the reduced expression of calprotectin was not dependent on TLR4-FA signalling in BMDM in vitro. The reduced expression of S100a8 and S100a9 likely indicates a reduced inflammatory capacity of BMM and myeloid precursors during NAFLD progression, possibly as a compensatory autocrine mechanism to limit local inflammation.30 31 Slpi was found to be strongly reduced in various cell clusters in the liver of WD-fed mice including MoMF and cDC. In line, a reduced expression of Slpi was also found in BMDM of WD-fed mice or FA-pretreated BMDM following LPS stimulation in vitro. Interestingly, SLPI has been reported to induce anti-inflammatory responses in human monocytes during APAP-induced acute liver failure32 and to inhibit TNF-α secretion.33 34 Furthermore, the reduced expression of neutrophil recruiting chemokines as Cxcl1, Cxcl2 and Cxcl3 in WD and FA-pretreated BMDM might also reflect a reduced inflammatory capacity of BMM. We did in our study not observe a direct effect on hepatic T cell recruitment following adoptive transfer of ND or WD BMM during APAP-induced liver injury. However, since we analysed the liver 12 hours after APAP, we cannot exclude an effect at later timepoints during injury, possibly related to the slower kinetics of T cell recruitment. This unique metabolic myeloid phenotype might aid in preventing hyperinflammation and on the other hand contribute to characteristic pathogenic consequences such as lipotoxicity, steatohepatitis and fibrosis.3 Metabolic reprogramming on WD feeding has been shown to depend on the modulation of the NLRP3 inflammasome by oxidised low-density lipoprotein, which is maintained even if the diet is switched back to low-fat feed.10 Our data suggest that oxidised low-density lipoprotein and free FA mediate innate immune reprogramming via TLR4-dependent signalling, thereby inducing unique changes in the inflammatory polarisation of myeloid immune cells.
One key observation of our study is that the persistent, obesity-related chronic inflammation prevented acute (excessive) inflammatory reactions of myeloid cells on acute injury to the liver. A reduced inflammatory capacity of the innate immune system related to WD feeding might serve as a negative feedback mechanism to limit inflammation during chronic injury without impairing innate immune defence against pathogens.10 In our study, sterile liver injury following APAP overdose was significantly attenuated in WD-fed mice, in line with previous studies.35 In full agreement with previous reports,36 we could also exclude potential differences in key enzymes of APAP metabolism due to NAFLD progression as a cause for attenuated liver injury. In line, while ND BMM increased APAP injury in WD mice, the adoptive transfer of WD BMM did not lead to an altered liver injury following APAP overdose in ND-fed mice, thereby underlining the substantial differences in the inflammatory potential of BMM from ND-fed compared with WD-fed mice during (sterile) liver injury. Future studies may investigate whether these metabolic alterations in the innate immune response are also relevant during infectious threats, for extrahepatic inflammation or for the surveillance of hepatic or extrahepatic malignancies.37
Our data underline the existence of a unique, common, FA-induced and TLR4-dependent immune cell polarisation in the liver and bone marrow in vivo, which has been previously suggested from in vitro work.7 26 The unique phenotype of myeloid cells in NAFLD is apparently already imprinted in myeloid bone marrow precursors but not in stem cells, and remains stable throughout extravasation, irrespective of changes in the local micromilieu or inflammatory stimuli. While our study focused on the consequences on hepatic myeloid cells and steatohepatitis, the adaptations, found in bone marrow and liver, are very likely also present in other extrahepatic tissues affected by metabolic syndrome, such as adipose tissue or vasculature. Our data imply that innate immunity is fundamentally changed as a consequence of Western-style, diet-induced obesity, which should be considered regarding therapeutic strategies aimed at modulating innate immune responses for treating obesity-related disease such as NASH, atherosclerosis, gut barrier dysfunction or type 2 diabetes.
Acknowledgments
We thank Sibille Sauer-Lehnen, Carmen Tag and Jasmin Hübner for their excellent technical assistance. We also thank Claus Neuhaus from the ZIEL - Institute for Food & Health (TU Munich, Germany) as well as Tom Clavel from the Institute of Medical Microbiology of the University Hospital of the RWTH Aachen for their support with the 16S rRNA analysis.
References
Footnotes
Contributors FT guided the research. OK designed, performed and analysed the animal and single-cell RNA sequencing experiments. JH performed the in vitro studies. ATA, DPPB and IGC performed the RNA sequencing analysis. JCM, MK, TR, TP, TL and CT contributed to research design and/or conducted the experiments. FT and OK wrote the manuscript. All authors reviewed and approved the manuscript.
Funding This work was supported by the German Research Foundation (DFG; Ta434/5-1 and SFB/TRR57). The study sponsor had no role in the study design or in the collection, analysis and interpretation of data.
Competing interests Work in the lab of FT has received research funding from Allergan, Galapagos, Inventiva and Bristol Myers Squibb.
Ethics approval All in vivo experiments were performed under conditions approved by the appropriate institutional and governmental authorities according to German legal requirements.
Provenance and peer review Not commissioned; externally peer reviewed.
Patient consent for publication Not required.