Atf3-mediated metabolic reprogramming in hepatic macrophage orchestrates metabolic dysfunction–associated steatohepatitis

Abstract

Metabolic dysfunction–associated steatohepatitis (MASH) is regulated by complex interplay between the macrophages and surrounding cells in the liver. Here, we show that Atf3 regulates glucose-fatty acid cycle in macrophages attenuates hepatocyte steatosis, and fibrogenesis in hepatic stellate cells (HSCs). Overexpression of Atf3 in macrophages protects against the development of MASH in Western diet–fed mice, whereas Atf3 ablation has the opposite effect. Mechanistically, Atf3 improves the reduction of fatty acid oxidation induced by glucose via forkhead box O1 (FoxO1) and Cd36. Atf3 inhibits FoxO1 activity via blocking Hdac1-mediated FoxO1 deacetylation at K242, K245, and K262 and increases Zdhhc4/5-mediated CD36 palmitoylation at C3, C7, C464, and C466; furthermore, macrophage Atf3 decreases hepatocytes lipogenesis and HSCs activation via retinol binding protein 4 (Rbp4). Anti-Rbp4 can prevent MASH progression that is induced by Atf3 deficiency in macrophages. This study identifies Atf3 as a regulator of glucose-fatty acid cycle. Targeting macrophage Atf3 or Rbp4 may be a plausible therapeutic strategy for MASH.

Metabolic dysfunction–associated steatohepatitis can be prevented by remodeling glucolipid metabolism in the liver macrophage.

INTRODUCTION

Metabolic dysfunction–associated steatohepatitis (MASH), characterized by severe hepatocellular steatosis, inflammation, and fibrosis, has emerged as the leading cause of cirrhosis, liver transplantation, and hepatocellular carcinoma. MASH is a progressive metabolic disease that displays dysregulation of glucolipid metabolism. The glucose–fatty acid cycle describes one set of mechanisms by which carbohydrate and fat metabolism interact. The destruction of glucose–fatty acid cycle contributes to the progression of MASH (1, 2). For example, malonyl–coenzyme A (CoA), as a key glucose-derived metabolite, is shown to disrupt fatty acid oxidation (FAO) by binding to carnitine palmitoyltransferase 1 (Cpt1) and inhibiting transport of cytosolic long-chain acyl-CoA molecules into the mitochondria for oxidation (3). Therefore, understanding the interplay between glucose metabolism and lipid metabolism and orchestrating MASH are critical concerns in this area of research. However, because of the heterogeneity and the complexity of glucolipid metabolism in liver cells, thus far, the underlying mechanisms of the interplay between glucose metabolism and lipid metabolism and its role in pathogenesis of MASH are not fully understood.

Many studies have demonstrated that cellular networks rather than a single cell type in the liver modulate MASH progression (4–6). Among liver cells, hepatic macrophages can orchestrate both the progression and restoration of MASH by cross-talk with surrounding cells, such as hepatocytes, hepatic stellate cells (HSCs), liver sinusoidal endothelial cells, and so forth (7–9). In addition, liver macrophages express high levels of glucose or lipid transporters, such as Glut3 and CD36, which functions as a pathway for not only glucose or free fatty acid (FFA) uptake and utilization but also regulation of macrophage functional state in inflammation and innate immunity (10, 11). Therefore, understanding the glucolipid metabolism networks of hepatic macrophage is critical for manipulation of the intercellular cross-talk between macrophages and their surrounding cells and development of appropriate therapeutic strategies for MASH.

In a clinical study, we found that activating transcription factor 3 (Atf3) protein expression in Cd68-positive liver macrophages is repressed in patients with MASH and there was a strong negative correlation between liver macrophage Atf3 protein levels and MASH score levels, indicating that Atf3 may be a critical regulator of liver macrophages in MASH progression. Atf3 is a member of the Atf/cAMP (cyclic adenosine monophosphate) response element–binding family of transcription factors. Our previous studies showed that hepatocyte Atf3 plays an important role in high-density lipoprotein (HDL) metabolism, bile acid metabolism, and triglyceride (TG) metabolism and prevents the progression of atherosclerosis and metabolic associated fatty liver disease (MAFLD) (12, 13). In addition, Atf3 mediates anti-inflammatory properties of HDL (14). However, the role of Atf3 in macrophage glucolipid metabolism and MASH is completely unknown. We found that overexpression of Atf3 could improve glucose-induced FAO reduction in liver macrophages during the development of MASH in mice. Therefore, we will examine the metabolic role of Atf3 in the liver macrophages in MASH progression.

In this context, we now show that Atf3 is a regulator of glucose–fatty acid cycle and Atf3 enhances macrophage FAO via reduction of cellular glucose levels by inhibition of forkhead box O1 (FoxO1)–mediated gluconeogenesis. In addition, Atf3 can also enhance FAO by induction of FFA uptake by Cd36. Mechanistically, Atf3 increases FoxO1 acetylation via blocking histone deacetylase 1 (Hdac1) deacetylating FoxO1 at K242, K245, and K262, resulting in reduction of gluconeogenesis and activation of AMP-activated protein kinase α (Ampkα)–mediated FAO. Moreover, Atf3 increases FFA uptake via palmitoylating CD36 at C3, C7, C464, and C466. Atf3 improves hepatic macrophage glucolipid metabolism, leading to the reduction of hepatocyte steatosis by inhibition of lipogenesis and the decrease in fibrogenesis in HSCs via retinol binding protein 4 (Rbp4). These findings provide an insight related to the metabolic role of liver macrophages that affects their surrounding cell function during MASH progression, suggest a plausible mechanism for the association of macrophage Atf3 with protection against hepatic steatosis and fibrosis in MASH, and indicate that Atf3 may be an effective liver macrophage–based therapeutic target for MASH.

RESULTS

Atf3 regulates glucose-induced FAO reduction in liver macrophages, and its expression is lower in patients with MASH or Western diet–fed mice

Thus far, the metabolic role of liver macrophages in MASH progression is not yet completely understood. In this study, we found that cellular glucose level was markedly higher in liver macrophages of mice fed on the Western diet (WD) for 24 weeks compared with those fed for 12 weeks or chow diet–fed mice (Fig. 1A, top blot), whereas FAO was lower in the mice fed WD for 24 weeks compared with those fed for 12 weeks or chow diet–fed mice (Fig. 1A, bottom blot). Atf3 expression was also notable reduced in liver macrophages of mice fed on the WD (Fig. 1B and fig. S1B). To investigate if Atf3 regulates glucose-induced FAO reduction in liver macrophages, we generated Atf3^fl/fl or Atf3^M−/− mice. Atf3^M−/− liver macrophages had increased glucose levels and decreased FAO, which could be normalized after Atf3 levels were recovered (Fig. 1, C to E). Moreover, Atf3 expression was lower in the livers or Cd68-positive hepatic macrophages of patients with MASH (Fig. 1, F to I), and there was a negative correlation between hepatic macrophage Atf3 expression and total MASH score (Fig. 1J). There was also a negative correlation between hepatic Atf3 mRNA expression and plasma glucose levels in humans (Fig. 1K). These data suggest that Atf3 may be a key regulator of glucose-reduced FAO in liver macrophages.

Fig. 1. Atf3 regulates glucose-induced FAO reduction in hepatic macrophages in WD-fed mice.

(A and B) C57BL/6J male mice were fed chow diet for 24 or 12 weeks followed by WD for an additional 12 or 24 weeks. Hepatic macrophages from the mice were isolated. Cellular glucose levels and FAO (A) and Atf3 protein (B) were determined (n = 3~5 per group). cpm, counts per minute. (C to E) Atf3^fl/fl or Atf3^M−/− liver macrophages were isolated and then infected with Lv-Atf3. After 24 hours, cellular glucose levels (C), FAO (D), and Atf3 protein (E) were measured. (F and G) Hepatic Atf3 protein and mRNA levels in humans with MASH were determined. (H) Atf3 mean fluorescence intensity in liver Cd68-positive cells from normal (n = 9 participants) or patients with MASH (n = 8 participants) was detected by immunostaining and then quantified. (I) Scale bars, 50 μm. (J) Correlation between liver macrophage Atf3 protein levels and total MASH score in humans. (K) Correlation between hepatic Atf3 mRNA levels and plasma glucose levels in humans. (L) The purification of isolated hepatic macrophages, stellate cells, or hepatocytes from mice were determined by qRT-PCR. *P < 0.05; **P < 0.01; NS, not significant. RLU, relative luciferase units.

Targeting liver macrophage Atf3 is sufficient to protect against WD-induced MASH

We have shown that liver macrophage Atf3 expression is markedly reduced in patients with MASH or WD-fed mice (Fig. 1H and fig. S1B). To study if liver macrophage Atf3 affects MASH progression, we transplanted Lv-Gfp or Lv-Atf3 macrophages into C57BL/6J mice and fed WD for 20 weeks. The transplanted macrophages expressing green fluorescent protein (GFP) are mainly located around the portal vein and in hepatic sinusoids (fig. S1C). As expected, compared with Lv-Gfp mice, the hepatic macrophage Atf3 protein levels in Lv-Atf3 mice were notable increased (Fig. 2A and fig. S1D). High expression of human Atf3 in macrophages did not alter plasma total cholesterol (TC) but slightly reduced liver–to–body weight ratio and plasma levels of aspartate aminotransferase (AST)/alanine aminotransferase (ALT), TG, and glucose (fig. S1, E to H). High expression of Atf3 reduced hepatic levels of TG, FFAs, and hydroxyproline by 36.2, 35.1, and 24.8%, respectively (Fig. 2, B to D). Histological studies confirmed that hepatic macrophage Atf3 overexpression decreased hepatic neutral lipid accumulation, fibrosis, and total MASH score (Fig. 2, E to G, and fig. S1I). Consistent with those findings, as shown in Fig. 2H, hepatic mRNA levels of genes involved in lipogenesis (Srebp1c, Acc, Fasn, and Scd1), M1 (inflammation) macrophage polarization (Tnfα, Il-1β, and Il-6), or fibrogenesis (Col1a1, Col1a2, Col3a1, and α-Sma) were decreased, and the mRNA levels of some genes involved in lipolysis and FAO (Hsl and Pdk4) were increased, while expression levels of some genes involved in M2 (anti-inflammatory) macrophage polarization (Arg1, Mrc1, and Ym1) were unchanged.

Fig. 2. Hepatic macrophage Atf3 protects against WD-induced steatohepatitis in mice.

(A to H) Lv-Gfp or Lv-Atf3 mice were fed WD for 20 weeks. After 20 weeks, the mice were euthanized. (A) Hepatic macrophage Atf3 proteins were analyzed by Western blot and then quantified (n = 4 per group). The levels of hepatic TG (B), hepatic FFA (C), and hepatic hydroxyproline (D) were measured (n = 7 per group). (E) Representative liver section images stained by hematoxylin and eosin (H&E) or picrosirius red (PSR). (F) MASH score. (G) Fibrotic area. (H) Hepatic mRNA levels. (I to P) Atf3^fl/fl or Atf3^M−/− mice were fed a WD for 20 weeks. After 20 weeks, the mice were euthanized. (I) Hepatic macrophage Atf3 protein levels. The levels of hepatic TG (J), hepatic FFA (K), and hepatic hydroxyproline (L) were determined (n = 7 per group). (M) Representative liver section images stained by H&E or picrosirius red. (N) MASH score. (O) Fibrotic area. (P) Hepatic mRNA levels. *P < 0.05; **P < 0.01.

To investigate if loss of macrophage Atf3 affected the development of MASH, we fed Atf3^fl/fl mice and Atf3^M−/− mice a WD diet for 20 weeks. Atf3^M−/− mice had higher TG levels by 28.1%, FFA levels by 36.4%, hydroxyproline levels by 56.6%, total MASH score levels by 24.1%, and fibrotic area by 35.3% in the liver (Fig. 2, I to O, and fig. S1O), but there was no change in plasma lipid levels (fig. S1M). Consistent with these data, Atf3^M−/− mice had increased hepatic mRNA expression levels of genes involved in lipogenesis (Srebp1c, Acc, Fasn, and Scd1), inflammation (Tnfα, Il-1β, and Il-6), or fibrogenesis (Col1a1, Col1a2, Col3a1, and α-Sma) and reduced mRNA expression levels of some genes involved in FAO, such as Pparα and Cpt2 (Fig. 2P). In addition, macrophage Atf3 deficiency increased the liver–to–body weight ratio and plasma levels of ALT/AST or glucose (fig. S1, J to L and N).

We fed Atf3^fl/fl mice and Atf3^M−/− mice an high-fat/cholesterol/fructose (HFCF) diet for 18 weeks, and similar results were collected (fig. S3, L to U). Together, our data demonstrate that targeting macrophage Atf3 is sufficient to protect against diet-induced MASH in mice.

Atf3 increases FAO via regulation of gluconeogenesis and fatty acid uptake

To investigate the mechanisms of Atf3 in regulation of glucolipid metabolism, we isolated liver macrophages from chow diet–fed Atf3^M−/− mice and did RNA sequencing (RNA-seq) to globally analyze gene expression. Many genes were differentially regulated by Atf3, with over 1000 genes being notable up-regulated or down-regulated by ≥1.5-fold [false discovery rate (FDR)–adjusted P value < 0.05]. Among the differentially regulated genes, only a limited number of genes were known to regulate glucolipid metabolism, such as Slc2a2, Pck1, Pck2, G6pc, Pdpr, Cpt2, and Ppara (Fig. 3A). We then did quantitative real-time polymerase chain reaction (qRT-PCR) and immunoblotting assays to confirm our RNA-seq data. Loss of Atf3 notable induced protein and mRNA levels of G6pc, Pck1, and Pck2 but reduced Ppara and CD36 protein and/or mRNA levels (Fig. 3, B and C). Consistent with the changes in gene expression, inactivation of Atf3 enhanced gluconeogenesis [phosphoenolpyruvate carboxykinase 2 (Pck2) activity] by 107.1% but attenuated fatty acid uptake or FAO by >30% (Fig. 3, D, F, and G), resulting in the accumulation of cellular glucose and TG in liver macrophages (Fig. 3E and fig. S2A). In contrast, overexpression of Atf3 reduced 48.9% of glucose levels and 33.1% of TG levels in liver macrophages via attenuated gluconeogenesis and induction of FAO (fig. S2, B to H). Atf3^M−/− liver macrophages had reduced FAO, which could be completely recovered after inhibition of gluconeogenesis by SKF-34288 hydrochloride (Fig. 3G, left). While it is true that the effects of Atf3 in regulation of FAO are abolished, SKF-34288 hydrochloride already has increased FAO (Fig. 3G, right).

Fig. 3. Atf3 enhances fatty acid utilization by inhibition of glucose-mediated AMPKα inactivation and up-regulation of Cd36.

(A) Hepatic macrophages were isolated from chow diet–fed Atf3^fl/fl or Atf3^M−/− mice, and total RNA was extracted for RNA-seq (n = 3). The heatmap shows that some glucolipid metabolism-related genes were up-regulated or down-regulated by Atf3. (B to G) Atf3^fl/fl or Atf3^M−/− mice were fed a WD for 12 weeks, and liver macrophages were isolated from the mice. (B) Relative mRNA levels of genes involved in glucolipid metabolism were quantified by qRT-PCR (n = 3 per group). Cell proteins were analyzed by Western blot assays and quantified (C). [(D) and (E)] Cellular Pck2 activity and glucose levels were measured. (F) Fatty acid uptake. (G) The Atf3^fl/fl or Atf3^M−/− hepatic macrophages were treated with 0.25 mM SKF-34288 hydrochloride or vehicle for 24 hours, and FAO was determined (left; n = 5 per group). The Lv-Gfp or Lv-Atf3 hepatic macrophages were treated with 0.25 mM SKF-34288 hydrochloride or vehicle for 24 hours, and FAO was determined (right, n = 4 per group). (H) Liver macrophages were isolated from C57BL/6J male mice, and CUT&Tag assay was performed. (I) Hepatic macrophages from Atf3^fl/fl or Atf3^M−/− mice were isolated and proteins were analyzed. (J and K) Hepatic macrophages were isolated from C57BL/6J mice and then pretreated with 10 μM BML-275 or dimethyl sulfoxide (DMSO). After 12 hours, the cells were infected with lentiviruses expressing Gfp (Lv-Gfp) or hAtf3 (Lv-Atf3) for 24 hours. FAO (J) and Pck2 activity (K) were measured. (L) Simplified model depicting the role of the ATF3 in regulating glucolipid metabolism. *P < 0.05; **P < 0.01. NS, not significant.

We next sought to understand how Atf3 induces FAO in liver macrophages. CUT&Tag assay was performed to detect the target genes of Atf3 in regulation of FAO. We did not find FAO-related genes (Pparα, Cpt2, and Cd36) that could be directly binding with Atf3 (Fig. 3H). It is noteworthy that Ampkα signaling can be regulated by Atf3. As shown in Fig. 3I, loss of Atf3 decreased protein levels of phosphorylated Ampkα (P-Ampkα) and phosphorylated acetyl-CoA carboxylase (P-Acc). In contrast, overexpression of Atf3 in liver macrophages increased P-Ampkα and P-Acc protein levels (fig. S2I). We reasoned that Ampkα may be a key signaling intermediate in this process as it is known that it could regulate FAO and serve as a sensor of glucose as well as cellular energy status (15, 16). Consistent with this idea, Atf3-induced FAO in hepatic macrophage was abrogated by inactivation of Ampkα via BML-275 (Fig. 3J). In addition, under the conditions of pretreatment of mouse hepatic macrophages with Ampkα inhibitor, Atf3 overexpression still reduces cellular Pck2 activity or gluconeogenesis (Fig. 3K). Data in Fig. 3 (A to K) indicate that Atf3 reduces cellular glucose levels by inhibition of gluconeogenesis, resulting in activation of Ampkα signaling and FAO in liver macrophages. The effects of Atf3 in regulation of cellular TG levels and FAO were also completely abolished after blocking fatty acid uptake in Cd36^−/− hepatic macrophages (fig. S3, A to C). After blocking CD36-mediated fatty acid uptake in liver macrophages, Atf3 did not alter the body weight, liver weight, plasma ALT, hepatic TG, FFA, or hydroxyproline but reduced plasma AST in WD-fed mice (fig. S3, D to J). Consistently, Atf3 did not regulate hepatic mRNA levels of genes involved in lipogenesis or fibrogenesis in the absence of macrophage Cd36 in these mice (fig. S3K). Together, these data demonstrate that Atf3 increases FAO by coordinated regulation of gluconeogenesis and fatty acid uptake (Fig. 3L).

Atf3 coordinates regulation of gluconeogenesis and fatty acid uptake via FoxO1 and Cd36

The findings suggest that Atf3 regulates FAO mainly by inhibition of gluconeogenesis (Fig. 3 and fig. S2), leading us to investigate the underlying mechanisms. The CUT&Tag data suggest that Atf3 regulates the expression of gluconeogenesis-related genes (G6pc, Pck1, and Pck2) indirectly (Fig. 3H). Atf3 protein could physically interact with FoxO1 in liver macrophages (Fig. 4, A and B, and fig. S4A), and the effects of Atf3 in regulation of gluconeogenesis, cellular glucose levels, and gluconeogenesis-related or FAO-related gene expression were all completely abolished after FoxO1 activity was inhibited by AS1842856 in hepatic macrophages (Fig. 4, C and D, and fig. S4, B to D). Acetylation modification of FoxO1 attenuates its transcriptional activity (17), and FoxO1 acetylation level in liver macrophages is reduced in WD-fed mice (Fig. 4E). Atf3 overexpression could increase FoxO1 acetylation in the liver of WD-fed mice or in the hepatic macrophages (Fig. 4, F and G), whereas inactivation of Atf3 decreased FoxO1 acetylation levels (Fig. 4F and fig. S4E). Thus, Atf3 reduces FoxO1 activity by induction of FoxO1 acetylation. Given that FoxO1 is also a key transcription factor that serves as a nutrient sensor for integrating insulin signaling to cell metabolism, we determined glucose/insulin resistance in HFCF-fed Atf3^M−/− mice as well as the control littermates (Atf3^fl/fl mice). Loss of Atf3 in macrophages worsens glucose intolerance but did not alter insulin resistance (fig. S3, V and W).

Fig. 4. Atf3 increases FoxO1 acetylation and CD36 palmitoylation.

(A) Co-immunoprecipitation (Co-IP) was performed using hepatic macrophages isolated from C57BL/6J mice fed a chow diet. (B) Representative image of basal FoxO1/Atf3 dimerization in hepatic macrophages. (C and D) Hepatic macrophages were pretreated with 1 μM AS1842856 for 24 hours and infected with Lv-Gfp or Lv-Atf3 for 24 hours. The Pck2 activity (C) and glucose levels (D) were determined. (E and L) C57BL/6J mice were fed chow diet (CD) or WD for 8 or 16 weeks. The levels of acetyl-FoxO1 and Cd36 palmitoylation in hepatic macrophages were measured. (F and M) Lv-Atf3, Atf3^M−/−, or control mice were fed WD for 20 weeks. The level of acetyl-FoxO1 and CD36 palmitoylation in liver macrophages was determined. (G) Hepatic macrophages were infected with Lv-Gfp or Lv-Atf3 for 36 hours. Protein samples were subjected to immunoprecipitation and Western blotting and developed with appropriate antibodies, as shown in the figures. (H) Co-IP was performed using hepatic macrophages isolated from C57BL/6J mice. (I) Hepatic macrophages were pretreated with 10 μM MS-275 for 24 hours and infected with Lv-Gfp or Lv-Atf3 for 24 hours. Protein samples were subjected to immunoprecipitation and Western blotting and developed with appropriate antibodies, as shown in the figures. (J) The FoxO1 has three putative Hdac1 deacetylating sites. (K) Human embryonic kidney (HEK) 293T cells cotransfected pCMV-Hdac1 or pCMV-Atf3 with pCMV-flag-FoxO1-Wt or pCMV-flag-FoxO1-Mut for 36 hours. Levels of acetylated FoxO1 were determined. (N) The levels of CD36 palmitoylation in Atf3^fl/fl or Atf3^M−/− macrophages were measured. (O) Hepatic macrophages were isolated from C57BL/6J mice, and CUT&Tag assay was performed. (P) The heatmap shows that palmitoylation-related genes were regulated by Atf3. (Q and R) Cd36^WT or Cd36^Mut macrophages were infected with Lv-Gfp or Lv-Atf3 for 24 hours. Fatty acid uptake and oxidation were measured. (S) Simplified model depicting the role of Atf3 in glucolipid metabolism. *P < 0.05; **P < 0.01. NS, not significant.

Next, to identify the intermediate that mediates Atf3 regulation of FoxO1 acetylation, we found that Hdac1 regulated FoxO1 acetylation in liver macrophages by physically interacting with FoxO1 protein (Fig. 4H and fig. S4F). We further examined if Hdac1 plays a role in Atf3-regulated FoxO1 acetylation, gluconeogenesis, or FAO. Atf3 regulated FoxO1 acetylation and gluconeogenesis-related or FAO-related gene expression in hepatic macrophages but not in MS-275 (an Hdac1 inhibitor) pretreated macrophages (Fig. 4I and fig. S4G). These data indicate that Hdac1 is important for Atf3 to regulate FoxO1 acetylation in liver macrophages. Last, we detected the possible Hdac1-mediated deacetylation sites of FoxO1. After screening, we found the three candidate acetylating sites (K242, K245, and K262) in the mouse FoxO1 proteins (Fig. 4J). Simultaneous mutations of three candidate acetylating sites (K242A, K245A, and K262A) completely abolished both the reduction of the mouse FoxO1 acetylation by Hdac1 and the induction of the FoxO1 acetylation by Atf3 (Fig. 4K). Collectively, the above data suggest that Atf3 enhances FoxO1 acetylation by blocking Hdac1-mediated deacetylation of the FoxO1 protein at K242, K245, and K262.

In addition, that Atf3 increases fatty acid uptake via CD36 in liver macrophages led us to ask how Atf3 up-regulates CD36 expression. The RNA-seq data and CUT&Tag indicate that Atf3 does not regulate CD36 expression at the transcriptional level (Fig. 3, A and H). CD36 palmitoylation plays an important role in CD36-mediated fatty acid uptake (18). We found that Atf3 increases CD36 palmitoylation levels via directly up-regulating palmitoylation-related enzymes, including abhydrolase domain containing 17A, depalmitoylase (Abhd17a), zinc finger DHHC-type containing 4 (Zdhhc4), and zinc finger DHHC-type containing 5 (Zdhhc5) in liver macrophages. Consistent with these data, ablation of Atf3 reduced CD36 palmitoylation levels and the gene expression of enzymes that mediate the CD36 palmitoylation (Fig. 4, N to P and fig. S5, A and B), whereas overexpression of Atf3 in macrophages had the opposite effect (fig. S5, C to F). Moreover, hepatic macrophage Atf3 overexpression could regulate CD36 palmitoylation in the liver of WD-fed mice (Fig. 4, L and M). To identify the palmitoylating sites of CD36 that mediates Atf3 regulation of fatty acid uptake, we found that Atf3 markedly increased CD36 palmitoylation, but the induction of CD36 palmitoylation mediated by Atf3 was abolished after simultaneous mutations of four palmitoylating sites (C3S, C7S, C464S, and C466S) (fig. S5, G and H). Mutation of palmitoylating sites of CD36 in liver macrophages has decreased fatty acid uptake by 38.4% and FAO by 37.2%, while overexpression of Atf3 in Cd36^Mut liver macrophages did not regulate fatty acid uptake and FAO (Fig. 4, Q and R). Together, data in Fig. 4 (L to R) and fig. S5 suggest that CD36 palmitoylation plays a critical role in Atf3 regulation of fatty acid uptake and FAO.

In sum, the data in Fig. 4 and figs. S4 and S5 supply mechanisms that Atf3 inhibits gluconeogenesis and enhances FAO by increasing FoxO1 acetylation and Cd36 palmitoylation, respectively (Fig. 4S).

Hepatic macrophage Atf3 inhibits hepatocyte lipogenesis and HSC activation mainly by Rbp4

The finding indicates that Atf3-enhanced FAO in macrophages attenuates hepatic steatosis and fibrosis, leading us to investigate how macrophage Atf3 regulated the function of hepatocytes and HSCs in the liver. We transplanted Lv-Gfp or Lv-Atf3 macrophages into C57BL/6J mice that were fed on a WD for 12 weeks, and then hepatocytes or HSCs were isolated from the mice. For hepatocytes, macrophage Atf3 expression led to a reduction in hepatocyte TG levels and newly synthesized TG levels but did not affect cellular TC levels (Fig. 5, A and B). Consistent with the phenotypes, the mRNA or protein levels of genes involved in hepatocyte lipogenesis (Srebp1c, Acc, and Fasn) were notable decreased, while expression levels of FAO-related genes (Cd36, Cpt1, and Cpt2) were not changed a lot (Fig. 5, C and D). In contrast, hepatocyte TG levels and newly synthesized TG levels were greatly potentiated by Atf3 ablation in macrophages (Fig. 5, E to H). For HSCs, high expression of Atf3 in macrophages markedly inhibited the expression of several fibrogenic genes in HSCs, such as Col1a1, Col1a2, Col3a1, and α-Sma (Fig. 6A), whereas loss of Atf3 in macrophages had opposite effects (Fig. 6B). To investigate if macrophage Atf3 can regulate HSC activation directly or indirectly by inhibition of hepatocyte lipid accumulation in vivo, the Atf3^M−/− mice were intraperitoneally injected with carbon tetrachloride (CCl₄) to induce liver fibrosis but had no significant impact on hepatocyte steatosis (fig. S6D). Loss of macrophage Atf3 increased hepatic hydroxyproline, fibrotic area, liver–to–body weight ratio, and plasma AST/ALT levels (fig. S6, E to I), leading to aggravated CCl₄-induced liver fibrosis in mice. These results suggest that macrophage Atf3 inhibits HSC activation directly in the liver. Thus, our data indicate that liver macrophage Atf3 simultaneously inhibits hepatocyte lipogenesis and HSC activation in mice.

Fig. 5. Liver macrophage Atf3 inhibits hepatocyte lipogenesis.

(A to D) Liver macrophages from C57BL/6J mice were isolated and then infected with Lv-Gfp or Lv-Atf3 for 24 hours. The cells were transplanted into C57BL/6J mice. After cell transplantation, the mice were fed a WD for 12 weeks. Newly synthesized TG was measured using GC-MS (B). Hepatocyte TG levels (A), mRNA levels (C), and protein levels (D) were measured. (E to H) Atf3^fl/fl or Atf3^M−/− mice were fed a WD. After 12 weeks, the newly synthesized TG was measured (F). Hepatocyte TG levels (E), mRNA levels (G), and protein levels (H) were determined. *P < 0.05; **P < 0.01. NS, not significant.

Fig. 6. Hepatic macrophage Atf3 inhibits hepatocyte lipogenesis and stellate cell activation mainly via Rbp4.

(A and B) Lv-Atf3, Atf3^M−/−, or control mice were fed a WD. After 20 weeks, HSCs from the mice were isolated and relative mRNA levels of genes involved in fibrogenesis were determined by qRT-PCR (n = 3 per group). (C and D) Hepatic macrophages from Atf3^fl/fl or Atf3^M−/− mice (fed a chow diet) were isolated, and total RNA was extracted for RNA-seq (n = 3). Some biological processes were regulated by Atf3 (C). The heatmap shows that some cytokines were up-regulated or down-regulated by Atf3 (D). (E) mRNA levels. (F) Protein levels. (G and H) Atf3^fl/fl or Atf3^M−/− mice were fed a WD. After 12 weeks, liver macrophages were isolated and cultured for 24 hours. Then, the media were collected and cocultured with freshly isolated hepatocytes (G) or HSCs (H) in the presence of anti-Rbp4 (10 mg/ml). After 12 hours, relative mRNA levels of genes involved in fibrogenesis were determined. (I and J) Liver macrophages from C57BL/6J mice were isolated and then infected with Lv-Gfp or Lv-Atf3 for 24 hours. The mRNA or protein levels of Rbp4 were measured. (K and L) Lv-Gfp or Lv-Atf3 liver macrophages were cultured for 24 hours. The media were collected and cocultured with primary hepatocytes (K) or mouse HSCs (L) in the presence of recombinant mouse Rbp4 protein or vehicle (5 μg/ml). The mRNA levels of hepatocytes or HSCs were determined. (M and N) Primary hepatocytes or mouse HSCs were infected with lentiviruses expressing shRNA against scramble sequences (Lv-shScr) or Stra6 (Lv-shStra6) for 24 hours and then incubated with recombinant mouse Rbp4 protein or vehicle (5 μg/ml). The mRNA levels of hepatocytes (M) or HSCs (N) were measured. *P < 0.05; **P < 0.01. NS, not significant.

Next, to investigate the underlying mechanisms of macrophage Atf3 in regulation of hepatocyte lipogenesis and HSC activation, we found that Rbp4 or fibroblast growth factor 21 (Fgf21) may be involved in macrophage Atf3 regulation of hepatocyte or HSC function (Fig. 6, C and D). Certain studies have reported that Rbp4 or Fgf21 are involved in MASH progression (19, 20). Atf3 deficiency markedly increased the mRNA or protein levels of Rbp4 in liver macrophages (Fig. 6, E and F), while Atf3 overexpression had the opposite effects (Fig. 6, I and J). Next, to determine if Rbp4 is essential for Atf3 reduction of hepatocyte lipogenesis and HSC activation, we isolated liver macrophages from Atf3^fl/fl or Atf3^M−/− mice. Conditioned media were collected from anti-Rbp4 antibody-treated Atf3^fl/fl macrophages or Atf3^M−/− macrophages and then added to hepatocytes or HSCs that were isolated from wild-type (Wt) C57BL/6J mice. As predicted, Atf3^M−/− macrophage media that induced the higher mRNA expression of lipogenic genes (Srebp1c, Fasn, Acc, and Scd1) in hepatocytes or fibrogenic genes (Col1a1, Col1a2, Col3a1, and α-Sma) in HSCs were all completely abolished by anti-Rbp4 antibody treatment (Fig. 6, G and H). Moreover, Rbp4 could block the reduction of lipogenic genes in hepatocytes or fibrogenic gene expression in HSCs by Lv-Atf3 macrophage media (Fig. 6, K and L). However, anti-Fgf21 could not block the reduction of fibrogenic gene expression by macrophage Atf3 overexpression (fig. S6A). Therefore, these findings suggest that macrophage Atf3 inhibits hepatocyte lipogenesis and HSC activation mainly by Rbp4.

Rbp4 is a secreted protein, and stimulated by retinoic acid 6 (Stra6) was established as a cellular receptor for Rbp4. To understand how Rbp4 regulates the expression of lipogenic genes in hepatocytes or fibrogenic genes in HSCs, we infected hepatocytes or HSCs with lentiviruses expressing Lv-shScr or Lv-shStra6. As expected, infection with Stra6 short hairpin RNA (shRNA) markedly reduced Stra6 expression. Inhibition of Stra6 also reduced lipogenic markers (Srebp1c, Fasn, Acc, and Scd1) in hepatocytes (Fig. 6M) or fibrogenic markers (Col1a1, Col1a2, Col3a1, and α-Sma) in HSCs (Fig. 6N). Consistent with the data of Fig. 6 (K and L), Rbp4 induces lipogenic gene expression in hepatocytes or fibrogenic gene expression in HSCs; however, Rbp4 failed to exert similar effects when Stra6 was inhibited (Fig. 6, M and N). These data demonstrate that Stra6 is required, at least in part, for Rbp4 to regulate lipogenic or fibrogenic gene expression in hepatocytes or HSCs. Rbp4 also did not regulate hepatocyte lipogenic gene expression in the absence of sterol regulatory element–binding transcription factor 1c (Srebp1c) (fig. S6B). Atf3^M−/− macrophage media that induced the higher expression of lipogenic genes (Fasn, Acc, and Scd1) could be abolished by anti-Rbp4 antibody treatment in Wt hepatocytes (Fig. 6G) but not in Srebp1c^−/− hepatocytes (fig. S6C). While it is true that the effects of anti-Rbp4 treatment are abolished, the Srebp1c inactivation already has decreased lipogenic gene expression.

Last, to understand how Atf3 inhibits Rbp4 expression in liver macrophages, we analyzed CUT&Tag data and found that Atf3 regulates Rbp4 expression indirectly (fig. S6J). We then treated liver macrophages with glucose, FFA mixtures, or vehicles. As shown in fig. S6 (K and L), glucose or FFAs can induce Rbp4 protein expression in liver macrophages, indicating that Atf3 reduces Rbp4 expression by mechanisms, at least partly, involving the role of Atf3 in lowering cellular levels of TG or glucose.

Targeting Rbp4 can prevent MASH progression induced by macrophage Atf3 inactivation in WD-fed mice

Above data indicated that Rbp4 may be involved in macrophage Atf3 regulation of MASH progression. Liver macrophage Rbp4 was also notable induced in WD-fed mice (Fig. 7I). Loss of macrophage Rbp4 prevents WD-induced MASH in mice by reduction of plasma ALT/AST, hepatic TG, FFA, hydroxyproline, MASH score, and fibrotic area (fig. S6, N to U). Analysis of hepatic mRNA levels by qRT-PCR showed that macrophage Rbp4 ablation reduced the expression of genes involved in lipogenesis (Srebp1c, Fasn, Acc, and Scd1), inflammation (Tnfα, Il-1β, and Il-6), or fibrogenesis (Col1a1, Col1a2, Col3a1, and α-Sma); however, there was no change in hepatic mRNA levels of genes involved in lipolysis (Hsl and Atgl) and FAO (Cpt1, Cpt2, Pdk4, and Cd36) (fig. S6V). Thus, loss of Rbp4 attenuates the development of diet-induced MASH. These findings led us to ask if Atf3 affects MASH after macrophage Rbp4 was ablated in vivo. To answer this question, we isolated hepatic macrophages from Rbp4^M−/− mice and then infected with Lv-Gfp or Lv-Atf3 to generate Rbp4^−/− liver macrophages carrying Gfp or Atf3. The macrophages were transplanted into C57BL/6J mice, and then the mice were fed a WD diet for 18 weeks. As expected, infection with Lv-Atf3 markedly increased Atf3 expression in liver macrophages in mice (Fig. 7A). Overexpression of Atf3 in Rbp4^−/− liver macrophages did not alter the body weight, liver weight, plasma ALT/AST, hepatic TG, FFA, or hydroxyproline in WD-fed mice (Fig. 7, B to G). Consistently, Atf3 also did not regulate hepatic mRNA levels of genes involved in lipogenesis or fibrogenesis in the absence of macrophage Rbp4 in these mice (Fig. 7H).

Fig. 7. Hepatic macrophage Atf3 protects against WD-induced MASH mainly by Rbp4.

(A to H) Hepatic macrophages from Rbp4^M−/− mice were isolated and then infected with Lv-Gfp or Lv-Atf3 for 24 hours. The macrophages were transplanted into C57BL/6J male mice and then fed a WD for 18 weeks (n = 6 per group). (A) Hepatic macrophage Atf3 protein levels. (B) Body weight and liver weight. (C) Plasma ALT/AST. (D) Hepatic lipid levels. (E) Hepatic FFA levels. (F) Hepatic hydroxyproline levels. (G) Representative liver section images stained by Oil Red O or picrosirius red. (H) Hepatic mRNA levels. (I) C57BL/6J mice were fed chow diet or WD fed for 8 or 16 weeks. Hepatic macrophages were isolated, and Rbp4 protein levels were determined. (J) Plasma Rbp4 protein levels of Lv-Atf3 mice or Atf3^M−/− mice were measured. (K to S) Atf3^M−/− mice were fed a WD for 13 weeks, and the mice were intraperitoneally (i.p.) injected with 100 μg of anti-Rbp4 antibodies (per mouse) or vehicle once every 6 days for 30 days (n = 6 per group). The experimental procedures are illustrated in (K). The levels of hepatic lipids (L), FFAs (M), hydroxyproline (N), and plasma ALT/AST were analyzed. (O) Representative liver section images stained by H&E or picrosirius red. (P) MASH score. (Q) Fibrotic area. (R) Plasma ALT/AST. (S) Hepatic mRNA levels. (T) Simplified model depicting the role of hepatic macrophage Atf3 in regulation of MASH in mice. *P < 0.05; **P < 0.01.

Rbp4 is essential for macrophage Atf3 in regulation of MASH progression. In vivo, hepatic macrophage Atf3 overexpression notable reduced plasma Rbp4 levels, while loss of Atf3 elevated plasma Rbp4 levels (Fig. 7J). These findings led us to ask if blocking of Rbp4 could prevent the progression of MASH in the Atf3^M−/− mice. Our data show that blocking Rbp4 could notable reduce the level of hepatic TG, FFA, hydroxyproline, MASH score, or fibrotic area but did not change hepatic TC levels and plasma ALT/AST levels in WD-fed Atf3^M−/− mice (Fig. 7, K to R). Hepatic mRNA or protein levels of genes involved in lipogenesis (Srebp1c, Acc, Fasn, and Scd1) or fibrogenesis (Col1a1, Col1a2, and α-Sma) were also reduced in these mice after blocking Rbp4 (Fig. 7S and fig. S6M). Together, data in Fig. 7 and fig. S6 demonstrate that Rbp4 is required for liver macrophage Atf3 to regulate hepatic steatosis and fibrosis. Targeting Rbp4 can prevent MASH progression induced by macrophage Atf3 ablation in mice.

DISCUSSION

Cellular glucose levels correlate negatively with FAO, indicating that there was a dysregulation of glucose–fatty acid cycle in liver macrophages during the progression of MASH. Given that the hepatic macrophage plays a critical role in regulating steatosis and fibrosis during MASH progression through cross-talk with hepatocytes or HSCs, targeting macrophage glucolipid metabolism may be an attractive approach for the therapy of dysfunctional glucolipid metabolism–related MASH. In this study, we provide compelling evidence demonstrating that macrophage Atf3 is a key regulator of glucose–fatty acid cycle, steatosis, inflammation, fibrosis, and MASH. Mechanistically, Atf3 inhibits gluconeogenesis via blocking Hdac1-mediated FoxO1 deacetylation at K242, K245, and K262, which results in reduction of cellular glucose levels and activation of Ampkα signaling and ultimately leads to the induction of FAO. Moreover, Atf3 can also enhance fatty acid uptake and FAO by increasing Zdhhc4/5-mediated CD36 palmitoylation. Last, Atf3-mediated metabolic reprogramming in macrophages notable attenuated hepatocyte lipogenesis and HSC fibrogenesis via Rbp4 in the liver (Fig. 7T).

FoxO1 is a key transcription factor involved in the modulation of gluconeogenesis and cellular glucose homeostasis (21). Myeloid FoxO1 depletion attenuates hepatic inflammation and prevents MASH (22). Acetylation is a reversible protein modification that inhibits FoxO1 transcriptional activity (17). Our findings identify Atf3 as a critical regulator of FoxO1 maintaining Ampkα-mediated FAO by inhibition of gluconeogenesis through binding with FoxO1 and blocking Hdac1-mediated deacetylation at K242, K245, and K262. Atf3-Hdac1-FoxO1 axis in hepatic macrophage plays a key role in regulating macrophage gluconeogenesis, Ampkα signaling, and FAO. Atf3-inhibited gluconeogenesis, reduced cellular glucose content, and enhanced FAO in liver macrophages were abolished by the inhibition of FoxO1 or Hdac1. In addition, considering that FoxO1 is a critical regulator in inflammation and insulin signaling, our data also indicate that macrophage Atf3 exerts an inhibitor effect on liver inflammation in WD-fed mice. However, macrophage Atf3 ablation did not alter the insulin resistance in these mice. It will be interesting to investigate why Atf3 that modulates FoxO1 is not involved in FoxO1-mediated regulation of insulin signaling. Therefore, our findings suggest that Atf3 might be an effective physiological target to selectively inhibit FoxO1 transcriptional activity and gluconeogenesis, resulting in improving glucose-induced FAO reduction and inflammation.

Maintaining FFA uptake helps to improve FAO. CD36 is shown as a high-affinity long-chain FFA transporter and facilitates FFA uptake (23, 24). Dynamic palmitoylation of CD36 can regulate FFA uptake. Thus far, two Zdhhc enzymes (Zdhhc4 and Zdhhc5) were found and could modulate the palmitoylation of CD36 at different subcellular localizations and FFA uptake (18). Here, we show that Atf3 increases CD36 palmitoylation levels at the cytoplasmic terminus of CD36 (C3S, C7S, C464S, and C466S) via inducing Zdhhc4/5 expression at the transcriptional level. On the basis of the data from Cd36^Mut hepatic macrophage or Cd36^−/− mouse models, it is likely that the major beneficial effects of Atf3 on FAO in liver macrophages are mechanisms involving the change of CD36 palmitoylation and the improvement of FFA uptake.

In addition, macrophage is a major source of cytokines, proinflammatory proteins, and profibrogenic growth factors, which can affect steatosis, inflammation, and fibrosis in the liver during MASH progression (7, 25). Our data indicated that Atf3 reprogrammed the metabolic role of liver macrophages, resulting in repressing the expression of Rbp4, which can directly inhibit hepatocyte lipogenesis and HSC fibrogenesis by binding to its membrane receptor Stra6. Rbp4 has been related to metabolic-related diseases, such as obesity and diabetes mellitus (26, 27). Elevating circulating Rbp4 levels are in relation to the risk of MASH (28). Here, we found that macrophage Atf3 attenuates MASH through Rbp4 signaling. Overexpression of Atf3 in Rbp4^−/− macrophages did not further improve hepatic steatosis and fibrosis in the WD-fed mice, while renormalized Rbp4 blunted the effects of Atf3 on modulation of Srebp1c-mediated hepatocyte lipogenesis and HSC activation. Combined with another interesting finding from this study that blocking of Rbp4 can prevent MASH progression in the WD-fed Atf3^M−/− mice, we can make a conclusion that Atf3 improved glucolipid metabolism of macrophage, leading to the reduction of Rbp4 expression, at least partly, involved in the critical protective role of Atf3 in MASH.

In summary, we have identified an intriguing link from Atf3 to FAO by Hdac1-mediated FoxO1 deacetylation or Zdhhc4/5-mediated CD36 palmitoylation, providing an insight into the mechanism of glucose-induced FAO reduction in liver macrophages during MASH progression. Moreover, our study finds the Atf3-Rbp4 axis as a critical pathway for targeting liver macrophage glucolipid metabolism to improve hepatic steatosis, inflammation, and fibrosis. As hepatocyte Atf3 is also important for maintaining lipid metabolism homeostasis in other settings, notably, atherosclerosis or liver steatosis, therefore, the induction of Atf3 pathway, specifically in liver macrophages, may be a promising therapeutic strategy for the treatment of MASH.

MATERIALS AND METHODS

Human tissues

Human liver tissues and paired plasma were collected at the Hubei University of Medicine Taihe Hospital, China. Normal liver tissues were collected from patients with hepatocellular hemangioma. Characteristics of participants are presented in table S1. All studies were approved by the institutional review board at Fudan University and the Hubei University of Medicine Taihe Hospital. Informed consent was obtained from human participants, and the collection of human tissues was approved by the institutional review board at the Hubei University of Medicine Taihe Hospital.

Mice and diets

C57BL/6J mice (stock no. N000013), Lyz2-cre mice (stock no. T003822), Atf3^fl/fl mice (stock no. T008416), and Rbp4^fl/fl mice (T010069) were purchased from GemPharmatech Co. Ltd. Floxed Atf3 mice or Floxed Rbp4 mice were then crossed with Lyz2-cre mice to generate Atf3^fl/fl Lyz2-cre mice (Atf3^M−/−) or Rbp4^fl/fl Lyz2-cre mice (Rbp4^M−/−). Hepatic macrophage Atf3 overexpression mice and control mice were generated by transplantation of 6 × 10⁶ hepatic macrophages that infected Lv-Gfp or Lv-Atf3 into C57BL/6J mice. Srebp1^−/− mice were purchased from the Jackson Laboratory. Cd36^−/− mice were kindly provided by T. Zhao (Fudan University, China). All the mice were housed in a temperature-controlled and humidity-controlled room with a 12-hour light–12-hour dark cycle and free access to food and water. Eight-week-old male mice were used unless otherwise specified. The mice were fed either a WD or an HFCF diet. The WD contained 42% fat and 0.2% cholesterol (Envigo, TD.88137). The HFCF diet contained 42% fat/0.2% cholesterol (Envigo, TD.88137) and 4.2% fructose (in drinking water). Mice were fed a WD or an HFCF diet for 18 to 24 weeks. Unless otherwise stated, all mice were fasted for 5 to 6 hours before euthanasia. All animal care and use procedures followed the guidelines of the Institutional Animal Care and Use Committee of Fudan University.

Chemicals

MS-275 (catalog no. HY-12163), SKF-34288 hydrochloride (catalog no. HY-128923), and BML-275 (catalog no. HY-13418A) were purchased from MedChemExpress. AS1842856 (catalog no. GC19040) was purchased from GLPBIO. CCl₄ (ZA051565) was purchased from SINOPHARM. ³H-palmitate (catalog no. NET043001MC) and ³H-oleate (catalog no. NET289001MC) were purchased from PerkinElmer. Recombinant mouse Rbp4 protein (catalog no. 3476-LC-050) and recombinant mouse FGF-21 protein (catalog no. 8409-FG-025/CF) were purchased from R&D Systems.

Plasmids and lentiviruses

To generate a lentiviral vector for human Atf3 overexpression, a 543–base pair (bp) cDNA fragment of hAtf3 was amplified by PCR and subcloned into a bicistronic lentiviral vector pLenti-CMV containing tagged Gfp (fig. S1A). The final hAtf3-lentiviral construct pLenti-hAtf3-GFP was verified by DNA sequencing. The lentivirus for expression of hAtf3 was named Lv-Atf3, and the control lentivirus was named Lv-Gfp. pLenti-Cd36^Wt lentivirus, pLenti-Cd36^Mut (C3S, C7S, C464S, and C466S) lentivirus, pLenti-shStra6 lentivirus, or pLenti-shScr lentivirus was generated and titrated by OBiO Technology Corp. (Shanghai, China). The plasmid containing open reading frame of mouse Cd36, Atf3, Hdac1, or FoxO1 was also purchased from OBiO Technology Corp. The mutants of cysteine in the cytoplasmic tail of CD36 (C3S, C7S, C464S, and C466S) or the mutants of FoxO1 acetylation sites (K242A, K245A, and K262A) were introduced by site-directed mutagenesis using QuikChange II Site-Directed Mutagenesis Kits.

Cell culture

Hepatic macrophages, primary hepatocytes, and HSCs were isolated from mice, as previously described (13, 29), and the purification of cells were detected by qRT-PCR (Fig. 1L). Hepatic macrophages from C57BL/6J male mice were isolated and cultured in Dulbecco’s modified Eagle’s medium (DMEM) containing vehicle (1 μM AS1842856 or 10 μM MS-275). After 12 hours, the macrophages were infected with lentiviruses [at a multiplicity of infection (MOI) = 75:1] for 12 hours, and mRNA levels were determined.

Hepatic macrophages were isolated from Atf3^M−/− mice and control mice and cultured in DMEM containing vehicle or FFA mixture (100 mM palmitate, 100 mM oleic acid, and 100 mM linoleic acid) for 24 hours. (i) The media were collected and cocultured with freshly isolated primary mouse hepatocytes. After 24 hours, the cellular lipid levels, fatty acid uptake, and oxidation of hepatocytes were determined. (ii) The media were collected and cocultured with freshly isolated primary mouse HSCs in the presence of anti-Rbp4 or anti-Fgf21 neutralization antibody (10 mg/ml). After 24 hours, mRNA levels in HSCs were determined.

Hepatic macrophages harboring the CD36 (C3S, C7S, C464S, and C466S) mutation (Cd36^Mut) were generated by infection of pLenti-Cd36^Mut (C3S, C7S, C464S, and C466S) into Cd36^−/− hepatic macrophage (MOI = 50:1). The control hepatic macrophages were generated by infection of pLenti-Cd36^Wt into Cd36^−/− hepatic macrophage.

Primary hepatocytes or mouse HSCs were infected with lentiviruses expressing shRNA against scramble sequences (Lv-shScr) or Stra6 (Lenti-shStra6) for 24 hours and then incubated with recombinant mouse Rbp4 protein or vehicle (5 μg/ml) for 12 hours. The mRNA levels of hepatocytes or HSCs were determined.

Hepatic macrophage transplantation in mice

Before cell transplantation, eight-week-old male C57BL/6J recipient mice were depleted residual hepatic macrophages by the tail vein injection of GdCl₃ (10 mg/kg) in saline, as previously described (30). Hepatic macrophages were isolated from male C57BL/6J donor mice and then infected with lentiviruses expressing Gfp or hAtf3 for 24 hours. For cell transplantation, macrophages (6 × 10⁶ cells per mouse) were injected into the portal vein in 200 μl of serum-free DMEM. After hepatic macrophage transplantation, the recipient mice were then fed a WD for 20 weeks.

Analysis of plasma or hepatic levels of lipids, hydroxyproline, AST, and ALT

Approximately 100 mg of liver tissue was homogenized in methanol, and lipids were extracted in chloroform/methanol (2:1 v/v), as previously described (31). TGs in the liver and plasma ALT (catalog no. TR71121) or AST (catalog no. TR70121) levels were measured using Infinity reagents from Thermo Fisher Scientific. Hepatic FFAs were measured using a kit from Wako Chemicals. Hepatic hydroxyproline (catalog no. STA675) or diacylglycerol (catalog no. MET-5028) levels were quantified using a kit from Cell Biolabs.

mRNA, qRT-PCR, and RNA-seq

Total RNA was extracted using TRIzol reagent (Thermo Fisher Scientific), and mRNA levels were quantified by qRT-PCR using PowerUP SYBR Green Master Mix (Thermo Fisher Scientific, A25778) on a 7500 Real-Time PCR machine (Applied Biosystems). The qRT-PCR program consisted of 1 cycle at 95°C for 5 min followed by 40 cycles at 95°C for 30 s, 60°C for 30 s, and 72°C for 30 s. The mRNA level of each gene was normalized to 36b4. The sequences of primers are shown in table S2.

To perform RNA-seq, total RNA was isolated from hepatic macrophages of chow diet–fed Atf3^fl/fl mice or Atf3^M−/− mice (n = 3 per group) using an miRVana miRNA Isolation kit (Thermo Fisher Scientific, catalog no. AM1560). cDNA libraries were constructed using the Illumina TruSeq Stranded Total RNA Library Prep kit (catalog no. 20020596) following the manufacturer’s instructions. RNA-seq was run on NextSeq 500 (Illumina). Sequencing data were processed in the BaseSpace Sequencing Hub. Briefly, sequenced reads were trimmed for adaptor sequences and mapped to the mouse mm10 whole genome using RNA-Seq Alignment (v2.0.2). Reads per kilobase transcript per million mapped reads and transcripts per million were collected using Illumina RNA-Seq Differential Expression (version 1.0.1). Genes were notable changed by >1.5-fold with an FDR-adjusted P value of <0.05. Volcano plot was generated using Qlucore Omics Explore 3.5 (Qlucore).

Protein extraction, immunoblotting, and immunoprecipitation

Western blot assays or immunoprecipitation were performed using whole liver lysates or cells lysates, as previously described (32, 33). Antibodies against Atf3 (catalog no. ABS136180), Cd36 (catalog no. ABS127810), or Srebp1c (catalog no. ABS131802) were purchased from Absin Bioscience Inc. Antibodies against Cd68 (catalog no. PE-65187), Hdac1 (catalog no. 10197-1-AP), P-Acc1 (catalog no. 29119-1-AP), glucose-6-phosphatase (G6pc; catalog no. 66860-1-lg), phosphoenolpyruvate carboxykinase 1 (Pck1; catalog no. 16754-1-AP), Pck2 (catalog no. 14892-1-AP), stearoyl-CoA desaturase 1 (Scd1; catalog no. 28678-1-AP), Abhd17a (catalog no. 15854-1-AP), Zdhhc5) (catalog no. 21324-1-AP), transforming growth factor–β1 (Tgfβ1; catalog no. 21898-1-AP), or α-tubulin (catalog no. 11224-1-AP) were purchased from Proteintech Group Inc. Antibodies against FoxO1 (catalog no. 14952S), acetylated lysine (catalog no. 9441), fatty acid synthase (Fasn; catalog no. 3180S), Acc (catalog no. 3662S), AMPKα (catalog no. 2603), or P- AMPKα (catalog no. 2535) were purchased from Cell Signaling Technology. Antibodies against Zdhhc4 (catalog no. ab235369) were purchased from Abcam. Antibodies against Rbp4 (catalog no. 1916AA0028) were purchased from MLBIO Biotechnology.

Proximity ligation assay and confocal immunofluorescence microscopy

Proximity ligation assay

Interactions between Atf3 and FoxO1 in hepatic macrophages of C57BL/6J mice were detected by in situ proximity ligation assay (PLA). PLAs were performed on fixed, permeabilized cell sections. The protocol for PLA was followed according to the manufacturer’s instructions (Olink Bioscience). The antibodies were used at the following concentrations: for Atf3, rabbit, catalog no. ABS136180, 1:200; for FoxO1, mouse, catalog no. 14952S, 1:200; PLA minus and PLA plus probes were added and incubated for 1 hour at 37°C. More oligonucleotides were then added and allowed to hybridize to the PLA probes. Ligase was used to join the two hybridized oligonucleotides into a closed circle. The DNA was then amplified (with rolling circle amplification), and detection of the amplicons was carried out using the Far-Red detection kit for fluorescence. Cell nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI). Slides were subsequently imaged using a multiple viewing microscope (Nikon, Japan).

Confocal immunofluorescence microscopy

To detect colocalization of Atf3 with FoxO1 in Kupffer cells, cells were cultured on poly-d-lysine–coated 25-mm round glass coverslips in six-well tissue culture plates overnight and then fixed with 4% paraformaldehyde at room temperature for 15 min. For immunostaining, cells were permeabilized with 0.1% Triton X-100 for 5 min and then blocked for 1 hour at room temperature. After washing, the cells were incubated overnight at 4°C with antibodies such as Atf3 antibodies, FoxO1 antibodies, or immunoglobulin G (IgG) antibodies diluted in blocking solution. Last, the coverslips were incubated with DAPI in blocking solution for 15 min at room temperature, washed with phosphate-buffered saline, drained, and mounted onto glass slides using Fluoromount-G (Fisher Scientific). Slides were subsequently imaged using an Olympus confocal microscope.

High-throughput CUT&Tag

CUT&Tag assay was performed as previously described with minor modifications (34). In brief, nuclei were purified from hepatic macrophages and incubated with concanavalin A–coated magnetic beads. The bead-bound cells were then resuspended with dig-wash buffer and a 1:50 dilution of primary Atf3 antibody (Absin, catalog no. ABS136180) or control IgG on a rotating platform overnight at 4°C. After washing in dig-wash buffer, beads were incubated with a secondary antibody. Then, the beads were washed in dig-Hisalt buffer and incubated with protein A–Tn5 transposome. Last, the cells were resuspended in a Tagmentation buffer, and the DNA fragments were purified by phenol-chloroform and used for sequencing. Sequencing was performed in the Illumina NovaSeq 6000 using 150-bp paired end following the manufacturer’s instructions.

Fatty acid uptake assay and FAO assay

Fatty acid uptake assay

To test the uptake of FFAs, the ³H-oleate uptake experiment was performed. Briefly, Lv-Atf3, Atf3^M−/−, or control hepatic macrophages were incubated in 100 μl of serum-free DMEM (5 × 10⁶ cells per reaction) and then incubated with 100 μl of assay buffer containing 40 μM oleate, 10 μM bovine serum albumin (BSA), and 0.05 μCi ³H-oleate at 37°C for 5 min. The reaction was stopped by adding 1 ml of cold stop buffer (500 μM phloretin and 0.1% BSA), and cells were pelleted at 1000g for 10 min, washed twice with stop buffer, resuspended in 100 μl of water, and analyzed by scintillation counting.

FAO assay

Hepatic macrophages were isolated and cultured in DMEM in 12-well dishes. FAO was performed using [³H]palmitate as a substrate, as previously described (35).

Hepatic lipogenesis assay

Hepatic lipogenesis was measured. In brief, mice were fasted for 4 hours and then intraperitoneally injected with ²H₂O (30 μl/g). After 4 hours, liver and plasma were snap-frozen in liquid nitrogen. The newly synthesized fatty acids or TGs were measured by gas chromatography–mass spectrometry (GC-MS) (Agilent 5977B Inert Plus MSC Turbo EI/CI Bundle), as previously described (36).

Cytological or histological staining

Cytological staining

Hepatocytes were fixed with 4% paraformaldehyde for 20 min and stained with Oil Red O (0.3%) for 5 min at 37°C. After nuclei were counterstained with hematoxylin, cellular lipid accumulation was observed under an Olympus confocal microscope.

Histological staining

Liver was fixed in 4% formalin and then embedded in optimal cutting temperature compound or paraffin. Hematoxylin and eosin (H&E),Oil Red O or Sirius Red staining was performed, as previously described (13). For immunohistochemical analysis of Atf3 and CD68, liver paraffin sections were stained with anti-Atf3 antibodies (Absin, catalog no. ABS136180) and anti-CD68 antibodies (Proteintech, catalog no. PE-65187). Species-specific secondary antibodies were covered sequentially. Images were captured using a microscope after nuclei were counterstained.

Statistical analysis

GraphPad Prism 8.4.2 was used for statistical analysis. Normality and equal variance tests were tested. If data passed both tests, unpaired Student’s t test followed by Welch’s corrections were performed. If the sample did not pass normality or equal variance test, statistical analyses were performed using unpaired Mann-Whitney U test. For more than two groups, one-way or two-way analysis of variance (ANOVA) with a post hoc Tukey’s or Dunnett’s test for multiple comparisons was performed. All values are expressed as means ± SEM. Differences were considered statistically significant at P < 0.05.

Supplementary Materials

This PDF file includes:

Figs. S1 to S6

Tables S1 and S2