Nuclear HMGB1 protects from nonalcoholic fatty liver disease through negative regulation of liver X receptor

Abstract

Dysregulations of lipid metabolism in the liver may trigger steatosis progression, leading to potentially severe clinical consequences such as nonalcoholic fatty liver diseases (NAFLDs). Molecular mechanisms underlying liver lipogenesis are very complex and fine-tuned by chromatin dynamics and multiple key transcription factors. Here, we demonstrate that the nuclear factor HMGB1 acts as a strong repressor of liver lipogenesis. Mice with liver-specific Hmgb1 deficiency display exacerbated liver steatosis, while Hmgb1-overexpressing mice exhibited a protection from fatty liver progression when subjected to nutritional stress. Global transcriptome and functional analysis revealed that the deletion of Hmgb1 gene enhances LXRα and PPARγ activity. HMGB1 repression is not mediated through nucleosome landscape reorganization but rather via a preferential DNA occupation in a region carrying genes regulated by LXRα and PPARγ. Together, these findings suggest that hepatocellular HMGB1 protects from liver steatosis development. HMGB1 may constitute a new attractive option to therapeutically target the LXRα-PPARγ axis during NAFLD.

Fig. 1.. Hepatocyte-specific Hmgb1-deleted mice on HFD or after fasting/refeeding challenge exhibit severe liver steatosis.

(A) Liver/body weight ratio, (B) Oil Red O staining on liver section with quantification, (C) neutral lipid analysis, and (D) mRNA expression of hepatic steatosis markers from liver biopsies of HMGB1^fl/fl and HMGB1^ΔHep mice subjected to 12-week HFD. (E) Liver/body weight ratio, (F) Oil Red O staining on liver section with quantification, (G) neutral lipid analysis, and (H) mRNA expression of hepatic steatosis markers from liver biopsies of HMGB1^fl/fl and HMGB1^ΔHep mice after a fasting/refeeding challenge. Data are means ± SEM from n = 7 (HMGB1^fl/fl) or n = 8 (HMGB1^ΔHep) per group for the HFD protocol (A to D) and from n = 8 (HMGB1^fl/fl) or n = 8 (HMGB1^ΔHep) per group for the F/R protocol (E to H). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by unpaired Mann-Whitney comparison.

Fig. 2.. Hepatic Hmgb1 overexpression decreases liver steatosis in mice subjected to HFD.

C57Bl6 mice were infected with either associated adenovirus expressing the eGFP (AAV8-CMV-GFP, n = 7) or Hmgb1 (AAV8-CMV-HMGB1, n = 9) sequence and then subjected to 12-week HFD regimen. (A) Immunoblot targeting HMGB1 and GFP in liver extracts. GAPDH was used as loading control. (B) Liver/body weight ratio. (C) Liver steatosis was determined by Oil Red O staining on liver sections, with the quantitative representation displayed on the right. (D and E) Neutral lipid analysis determined by MS (D) and mRNA expression of hepatic steatosis markers analyzed by RT-qPCR (E). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by unpaired Mann-Whitney comparison. a.u., arbitrary units.

Fig. 3.. Hmgb1 deletion increases hepatocyte lipid synthesis in vitro and in vivo.

(A) In vivo, lipogenesis was measured on HMGB1^fl/fl (n = 5) and HMGB1^ΔHep (n = 5) mice. Mice were food-deprived for 6 hours and then injected with ³H-glucose (0.4 μCi/g, intraperitoneally) and euthanized 1 hour later, and ³H was measured in TG fraction of liver and adipose tissues [perigonadial adipose tissue (PG), subcutaneous adipose tissue (SC), and BAT]. (B to D) HMGB1^fl/fl (n = 5) and HMGB1^ΔHep (n = 5) mice were treated for 3 weeks with D20 in the drinking water (4%), under CD (B), upon HFD60% (C), and after F/R challenge (D). Liver lipogenesis was determined by the amount of ²H incorporated in palmitate normalized by ²H-labeled plasmatic water. (E and F) In vivo, assessment of liver lipoprotein secretion was determined by measuring circulating triacylglycerol concentration (n = 4 per genotype and diet) (E) and liver MTP activity (F), HMGB1^fl/fl (n = 4) and HMGB1^ΔHep (n = 5). (G and H) Lipid synthesis was measured in vitro, on primary hepatocytes isolated from adult HMGB1^fl/fl (n = 7 to 9) and HMGB1^ΔHep (n = 8 to 9) mice on CD (G) and HFD (H). Data are means ± SEM of three independent experiments. *P < 0.05 by unpaired Mann-Whitney comparison or two-way ANOVA. ^$P < 0.05 and ^$$P < 0.01, for treatment effect by one-way ANOVA.

Fig. 4.. Microarray analysis of hepatic gene expression profiles in HMGB1^ΔHep mice.

(A) Heatmap showing genes that are differentially expressed in the livers of HMGB1^ΔHep mice compared to HMGB1^fl/fl mice (fold change > 1.5; P ≤ 0.01) after HFD (left) or F/R (right). Heatmaps display the mean normalized expression per genotype per nutritional challenge. (B) Venn diagram displaying overlap between up- and down-regulated genes in the two regimens. (C) Heatmap displaying only differentially expressed genes commonly found in both regimens (fold change > 1.5; P ≤ 0.01). (D and E) Top 5 GO biological processes enriched using gene sets for each regimen, with the −log₁₀(P value) of enrichment shown as bars and the number of matched genes as colored lines. (F) Network displaying Reactome pathways related to metabolism that are enriched by our HMGB1 gene sets from both nutritional challenges. Edge thickness represents the number of genes regulated by HMGB1 among each subcategory. (G) Top upstream regulators identified using the ChEA database, with the −log₁₀(P value) of enrichment as bars and the number of matched genes as the green line. Data are means ± SEM from n = 4 (HMGB1^fl/fl) or n = 4 (HMGB1^ΔHep) per group for the 12-week HFD protocol and from n = 4 (HMGB1^fl/fl) or n = 4 (HMGB1^ΔHep) per group for the F/R protocol.

Fig. 5.. In vivo knockdown of LXR and PPARγ normalizes liver steatosis in HMGB1^ΔHep mice.

(A and B) HMGB1^fl/fl [vehicle (Veh), n = 5; T0901317 (T09), n = 10] and HMGB1^ΔHep (Veh, n = 3; T09, n = 7) mice were treated with either vehicle (5% carboxymethylcellulose) or LXR synthetic agonist T0901317 (oral gavage, 30 mg/kg per day) for four consecutive days. After 6-hour starvation on the last day, mice were sacrificed. (A) Liver tissue was then subjected to RT-qPCR analysis of the indicated LXR-dependent genes. (B) HMGB1^fl/fl (n = 10) and HMGB1^ΔHep (n = 12) mice were infected with either adenovirus expressing an LXR shRNA or a scramble (SCR) sequence and then subjected 7 days later to an F/R challenge. Liver steatosis was determined by Oil Red O (ORO) staining on liver sections, with the quantitative representation displayed on the right. (C) HMGB1^fl/fl [Veh, n = 4 to 6; rosiglitazone (Rosi), n = 7] and HMGB1^ΔHep (Veh, n = 4 to 6; Rosi, n = 8) mice were treated with either vehicle [5% dimethyl sulfoxide (DMSO) in saline] or PPARγ synthetic agonist rosiglitazone (30 mg/kg per day, intravenously) and were sacrificed 18 hours later. Liver tissue was then subjected to RT-qPCR analysis of the indicated PPARγ-dependent genes. (D) HMGB1^fl/fl (ShSCR, n = 7; ShPPARγ, n = 8) and HMGB1^ΔHep (ShSCR, n = 8; ShPPARγ, n = 7) mice were infected with either adenovirus expressing a PPARγ shRNA or a scramble (SCR) sequence and then subjected 7 days later to an F/R challenge. Liver steatosis was determined by Oil Red O staining on liver sections. Data are means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, HMGB1^fl/fl and HMGB1^ΔHep comparison, by unpaired Mann-Whitney comparison. ^$P < 0.05 and ^$$P < 0.01, for treatment effect by one-way ANOVA.

Fig. 6.. ChIP-seq identified a subset of LXR-responsive genes to be negatively regulated by HMGB1 during liver steatosis.

(A) PCA score plot of ChIP-seq data of liver tissue from HMGB1^fl/fl mice on CD (green) or subjected to F/R (red) or HFD (purple). (B) Venn diagram showing the number of HMGB1 binding peaks, (C) UCSC genome browser of tracks (stacked) showing HMGB1 differential chromatin occupancy, and (D) average signal density profiles around transcription starting site in different nutritional states: CD (green) or during HFD (purple) or after F/R (red). (E and F) Functional enrichment analyses showing GO terms associated with the differential HMGB1 chromatin binding sites between (E) CD and HFD and (F) CD and F/R. (G) Venn diagram displaying shared enriched genes (n = 134) displaying a very high occupancy rate during fed state belonging to “Integration of energy metabolism” and “Phospholipid metabolism” GO functions compared to HFD (purple) and F/R (red). (H) Bar graph displaying consensus motifs in promoters of the 134 genes differentially occupied by HMGB1 via OPOSUM analysis; the bars represent the z score. (I and J) Heatmaps displaying the mean microarray expression levels for the 134 genes identified by ChIP-seq in liver from HMGB1^fl/fl (n = 4) and HMGB1^ΔHep (n = 4) mice subjected to either HFD (I) or F/R (J).

Fig. 7.. HMGB1 represses LXRα, but not PPARγ, transcriptional activity in vitro.

(A) Effect of HMGB1 on LXRE-luciferase reporter activity. Ad293 cells were treated with DMSO (vehicle), T0901317 (T09) (0.1 μM), and/or LG286 (1 nM) for 14 hours. (B) Effect of HMGB1 on PPRE (PPAR responsive element)-luciferase reporter activity. Ad293 cells were treated with DMSO (vehicle) and rosiglitazone (Rosi; 1 μM, overnight). (C) Co-IP assay was performed to detect a potential interaction between HMGB1 and LXR in Ad293-transfected cells treated with DMSO (vehicle) or T0901317 (0.1 nM for 14 hours). Data are representative of three independent experiments. (D and E) IGV (Integrative Genomics Viewer) genome browser shot of HMGB1 and LXRα ChIP-seq data along the locus of Acly (D) and Fasn (E) gene loci. HMGB1 tracks in liver from HMGB1^fl/fl upon CD (green), upon HFD (purple), and after F/R (red). LXRα tracks from basal liver (dark orange) and T0901317 challenged liver (light orange). Gene loci displayed in gene model (blue) are displayed on the bottom track. (F) Gene expression of direct (Srebf1, Scd-1, Abcg-5, and Abcg-8) and indirect (Cd-36, Cidec, Pnpla3, and Fasn) targets of LXRα in livers of HMGB1^fl/fl (n = 7) and HMGB1^ΔHep (n = 9) mice. (G) Adult HMGB1^fl/fl mice were infected with either AAV8-TBG-GFP (n = 8) or AAV8-TBG-Cre (n = 9) to selectively generate Hmgb1 deletion in hepatocytes in vivo, and expression of direct (Srebf1, Scd-1, Abcg-5, and Abcg-8) and indirect (Cd-36, Cidec, Pnpla3, and Fasn) responsive genes was determined using RT-qPCR. Data are means ± SEM of three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by unpaired Mann-Whitney comparison or two-way ANOVA. ^$P < 0.05, ^$$P < 0.01, and ^$$$P < 0.001, for treatment effect by two-way ANOVA.