REDOX-SENSITIVE HIGH-MOBILITY GROUP BOX-1 ISOFORMS CONTRIBUTE TO LIVER FIBROSIS PROGRESSION AND RESOLUTION IN MICE

Abstract

Background & Aims:

high-mobility group box-1 (HMGB1) significantly increases and undergoes post-translational modifications (PTMs) in response to liver injury. Since oxidant stress plays a major role in liver fibrosis and induces PTMs in proteins, we hypothesized that redox-sensitive HMGB1 isoforms contribute to liver fibrosis progression and resolution.

Methods & Results:

using electrospray ionization-liquid chromatography-mass spectrometry (ESI-LC-MS), we identified that disulfide ([O]) and sulfonated ([SO₃]) HMGB1 rise during CCl₄-induced liver fibrosis progression, however while [O] HMGB1 declines, [SO₃] HMGB1 drops but remains, during fibrosis resolution. Conditional knockout mice of Hmgb1 revealed that production of [O] and [SO₃] HMGB1 occurs mostly in hepatocytes. Co-injection of [O] HMGB1 worsens CCl₄-induced liver fibrosis more than co-injection of [H] HMGB1. Conversely, ablation of [O] Hmgb1 in hepatocytes reduces liver fibrosis. Moreover, ablation of the receptor for advanced-glycosylation end-products (Rage), reveals that the pro-fibrogenic effect of [O] HMGB1 occurs through RAGE signaling in hepatic stellate cells (HSCs). Notably, injection of [SO₃] HMGB1 accelerates fibrosis resolution due to RAGE-dependent stimulation of HSC apoptosis. Importantly, gene signatures activated by redox-sensitive HMGB1 isoforms in mice classify fibrotic patients according to fibrosis and inflammation scores.

Conclusion:

dynamic changes in hepatocyte-derived [O] and [SO₃] HMGB1, signal through RAGE-dependent mechanisms on HSCs, to drive their pro-fibrogenic phenotype and fate, contributing to progression and resolution of liver fibrosis.

GRAPHICAL ABSTRACT

INTRODUCTION

Liver fibrosis, a common manifestation of chronic liver disease, is a dynamic process that can be lessened by slowing progression and/or promoting resolution. Despite recent advances in therapies for chronic hepatitis, unfortunately fibrosis remains in many patients, posing a major challenge for survival [1, 2]. Therapeutic strategies for liver fibrosis are lacking due to incomplete understanding of fibrosis progression and resolution. Thus, identifying underlying molecular mechanisms is vitally important, for rapid and effective targeting of signals driving the HSC pro-fibrogenic phenotype and fate, perhaps with direct anti-fibrotics and/or inducers of activated HSC clearance or deactivation.

HMGB1 is a damage-associated molecular pattern increased in response to sterile and non-sterile tissue injury [3]. While initially believed to be simply an architectural protein [4], increased nuclear export due to injury, drives HMGB1 cytoplasmic localization and secretion [3]. Recent studies from our group showed that HMGB1 is secreted largely by hepatocytes, and to some degree, by Kupffer cells and infiltrating macrophages [3, 5]. Consequently, HMGB1 can communicate with, and amplify injury to neighboring cells, including HSCs, through receptor activation, since they express RAGE [5], a key HMGB1 receptor [5, 6].

Oxidant stress plays a major role in liver fibrosis [7], and induces PTMs in proteins modifying their function [8]. So far, the role of oxidative modifications of HMGB1 in liver fibrosis progression and resolution is unknown. We hypothesized that redox-sensitive HMGB1 isoforms contribute to liver fibrosis progression and resolution. We found that dynamic changes in [O] and [SO₃] HMGB1 signal through RAGE-dependent mechanisms on HSCs, to drive their pro-fibrogenic phenotype and fate, contributing to progression and resolution of liver fibrosis. Therefore, these redox-sensitive HMGB1 isoforms could be considered biomarkers of disease progression and resolution, and prospective therapeutic targets.

MATERIALS AND METHODS

Ethical guidelines.

All animals received humane care according to the criteria outlined in the Guide for Care and Use of Laboratory Animals. Housing, breeding, and husbandry conditions were approved by the University of Illinois at Chicago IACUC office prior to initiating the studies. All in vivo experiments were carried out according to the Animal Research: Reporting of In Vivo Experiments guidelines.

Statistical methods.

All data are expressed as mean ± SEM. Statistical comparisons between groups and in vitro treatments were performed using paired Student’s t-test. All in vitro experiments were performed in triplicate, and representative blots are shown in all figures. All in vivo data were analyzed by two-factor ANOVA. The number of mice is indicated in the legend to figures. Both sexes were combined in the statistical analysis as no significant differences were found between them. P values are indicated on the figure legends.

Detailed methods are described in the supplementary data file.

RESULTS

Hmgb1 ablation in hepatocytes and myeloid cells facilitates fibrosis resolution.

We previously demonstrated that Hmgb1 ablation in hepatocytes and myeloid cells (mostly Kupffer cells and infiltrating macrophages) partially protects, while ablation in both, prevents carbon tetrachloride (CCl₄)-induced liver fibrosis [5]. Similar protection occurred in the thioacetamide and bile duct ligation models of liver fibrosis [5]. To establish if hepatocyte and/or myeloid cell-derived HMGB1 contributes to fibrosis resolution, WT, Hmgb1^ΔHep, Hmgb1^ΔMye and Hmgb1^ΔHepΔMye mice, were injected mineral oil (MO) or CCl₄. Half of them were sacrificed after 1 month (fibrosis progression), and the other half, 1 week after ceasing MO or CCl₄ injections (fibrosis resolution).

Hematoxylin and eosin (H&E) staining, HMGB1 and COL1 immunohistochemistry (IHC) (Fig. 1A top), F4/80 and TNFA IHC and quantification (Fig. S1A top, S1B top), the pathology scores (Fig. 1B top), HMGB1 morphometry (Fig. S1C top), COL1 quantification by morphometry and western blot (Fig. 1B top right, 1C left), and alanine aminotransferase (ALT) activity (Fig. 1D left), revealed partial protection from liver fibrosis in CCl₄-injected Hmgb1^ΔHep and Hmgb1^ΔMye, while almost full prevention in Hmgb1^ΔHepΔMye, compared to WT mice. The endothelial cell marker CD31, shown by IHC (Fig. S1D top) and quantified by morphometry (Fig. S1D bottom), was alike in all groups of CCl₄-injected mice. Notably, we observed modest resolution from CCl₄-induced liver fibrosis in WT, partial in Hmgb1^ΔHep and Hmgb1^ΔMye, however it was nearly full in Hmgb1^ΔHepΔMye mice (Fig. 1A–1B bottom, 1C–1D right, Fig. S1A bottom, S1B bottom). These findings suggest that Hmgb1 ablation in hepatocytes or myeloid cells partially protects, whereas ablation in both prevents CCl₄-induced liver fibrosis, and facilitates fibrosis resolution.

Fig. 1. Hmgb1 ablation in hepatocytes and myeloid cells facilitates fibrosis resolution.

WT, Hmgb1^ΔHep, Hmgb1^ΔMye and Hmgb1^ΔHepΔMye mice were injected CCl₄ for 1 month (progression). Following cessation of CCl₄ treatment, mice were allowed to recover for 1 week (resolution). (A top) H&E, HMGB1 and COL1 IHC, in 1-month CCl₄-injected WT show more necrosis (blue arrows), inflammation (yellow arrows), HMGB1 expression and translocation into the cytoplasm or cell targeting efficiency (green arrows), perivenular fibrosis (closed pink arrows), and bridging fibrosis (open pink arrows) than in Hmgb1^ΔHep, Hmgb1^ΔMye or Hmgb1^ΔHepΔMye mice. (A bottom) H&E, HMGB1 and COL1 IHC, at 1 week of resolution from CCl₄-induced fibrosis in WT, shows more inflammation and fibrosis than in Hmgb1^ΔHep, Hmgb1^ΔMye or Hmgb1^ΔHepΔMye mice. (B) Necrosis, hepatocyte ballooning degeneration, inflammation, and fibrosis scores, and COL1 morphometry. (C) Western blot analysis for COL1. Results are expressed as fold-change of CCl₄-injected WT mice (progression) assigned a value of 1. (D) Serum ALT activity. (E) HMGB1 protein structure with Cys 23, 45, and 106 that can undergo oxidation written in red (F). Identification of redox-sensitive HMGB1 isoforms and their cellular source, in CCl₄-induced liver fibrosis progression and resolution. Quantification of HMGB1 isoforms ([H], [O], [SO₃]) in liver (ng/mg) by ESI-LC-MS. In all panels, results are expressed as mean ± SEM; n=6/group; *p<0.05, **p<0.01 and ***p<0.001 vs. WT + CCl₄; •p<0.05 and ••p<0.01 vs. WT + CCl₄ resolution.

Identification of signature redox-sensitive HMGB1 isoforms and their cellular source, in CCl₄-induced liver fibrosis progression and resolution.

Since oxidant stress is a major driver of liver fibrosis, and the antioxidant S-adenosyl methionine (SAMe) lowered HMGB1 (Fig. S2) and protected from fibrosis [9], we asked whether oxidative modifications of HMGB1 could occur, and play a role in progression and/or resolution of liver fibrosis. First, we measured total HMGB1 by ELISA to find that during liver fibrosis progression and resolution, there was significant decrease in liver HMGB1 in Hmgb1^ΔHep and Hmgb1^ΔHepΔMye, and in serum HMGB1 in Hmgb1^ΔHepΔMye, compared to WT mice (Fig. S3).

To dissect whether HMGB1 undergoes oxidation, identify the cellular source of the potential redox-sensitive HMGB1 isoforms, and quantify their dynamic changes, we analyzed it in the livers from WT, Hmgb1^ΔHep, Hmgb1^ΔMye and Hmgb1^ΔHepΔMye mice during fibrosis progression and resolution, using electrospray ionization-liquid chromatography-mass spectrometry (ESI-LC-MS). A schematic representation of the HMGB1 protein structure with cysteines (Cys) 23, 45 and 106, that can undergo oxidation is shown in Fig. 1E. During fibrosis progression, HMGB1 underwent oxidation to disulfide ([O]) and sulfonated ([SO₃]) HMGB1, with the latter being irreversible. Hepatocytes were the main source of [O] and [SO₃] HMGB1, as these isoforms decreased in liver from CCl₄-injected Hmgb1^ΔHep or Hmgb1^ΔHepΔMye, compared to WT mice (Fig. 1F top). During fibrosis resolution, [O] and [SO₃] HMGB1 were present in all groups of mice, albeit at lower concentration than in fibrosis progression, except for [SO₃] HMGB1 that remained elevated (Fig. 1F bottom). In serum, these isoforms were below detection limit. Nevertheless, additional experiments were needed to define the role of each isoform during fibrosis progression and resolution.

Co-injection of [O] HMGB1 increases more CCl₄-induced liver fibrosis than co-injection of [H] HMGB1.

To dissect if the HMGB1 isoforms drive liver fibrosis progression, WT mice were injected [H] or [O] HMGB1 [10], before each CCl₄ dose, for 1 month. HMGB1 morphometry showed similar HMGB1 expression in all mice (Fig. S4A). Liver injury and fibrosis, assessed by H&E staining and IHC for HMGB1, COL1 and DES (HSC marker) (Fig. 2A top four rows), the pathology scores (Fig. 2B left), quantification of COL1 by morphometry and western blot (Fig. 2B right, 2C), and ALT activity (Fig. 2D), increased more in mice co-injected [O] than [H] HMGB1, compared to mice injected CCl₄ alone. Because of the key role of macrophages in liver fibrosis [11], we performed IHC for F4/80 (a marker of these cells), and TNFA, known to trigger hepatocyte injury [12]. Co-injection of [H] HMGB1 increased the number of F4/80⁺ and TNFA⁺ cells, yet it was greater by co-injection of [O] HMGB1, compared to injection of CCl₄ alone (Fig. 2A bottom two rows). These results suggest a dual mode of action of the HMGB1 isoforms, whereby [H] HMGB1 increases HSC migration and proliferation, and [O] HMGB1 enhances inflammation and HSC activation, both of which contribute to fibrosis progression.

Fig. 2. Co-injection of [O] HMGB1 increases more CCl₄-induced liver fibrosis than co-injection of [H] HMGB1.

WT mice were injected i.p. 0–0.1 μg/g of endotoxin-free [H] or [O] HMGB1, 1 hour before each CCl₄ dose, for 1 month. (A) H&E staining showing necrosis (blue arrows) and inflammation (yellow arrows). IHC for HMGB1 (green arrows), COL1 showing perivenular fibrosis (closed pink arrows) and bridging fibrosis (open pink arrows), DES (green arrows), F4/80 (green arrows), and TNFA (green arrows). (B) Necrosis, hepatocyte ballooning degeneration, inflammation, and fibrosis scores, and COL1 morphometry. (C) Western blot analysis for COL1. Results are expressed as fold-change of control assigned a value of 1. (D) Serum ALT activity. (E) Pathway analysis from whole liver RNAseq data using IPA. Red and blue represent positive or negative Z-scores, indicating positive or negative regulation of the pathway, respectively (pink arrows: profibrogenic pathways). (F) GSEA from RNAseq data showing increased signatures of collagen synthesis and markers of fibroblasts in WT mice co-injected [O] HMGB1, compared to mice injected CCl₄ alone. In all panels, results are expressed as mean ± SEM; n=4/group; *p<0.05, **p<0.01 and ***p<0.001 co-injected vs. CCl₄.

We then performed RNA sequencing (RNAseq), ingenuity pathway analysis (IPA), and gene set enrichment analysis (GSEA), and found increased profibrogenic pathways (hepatic fibrosis, HSC activation, idiopatic pulmonary fibrosis), and enrichment in profibrogenic reactomes (collagen formation, smooth muscle contraction, collagens), in WT mice co-injected [O] HMGB1, but not in mice co-injected [H] HMGB1 (not shown for the reactomes), compared to mice injected CCl₄ alone (Fig. 2E–2F). Thus, [O] enhances liver fibrosis more than [H] HMGB1, yet how this occurred, and whether these HMGB1 isoforms could be potential therapeutic targets to prevent fibrosis progression, remains unknown.

Overexpression of WT HMGB1 in hepatocytes promotes, whereas ablation of [O] HMGB1 reduces CCl₄-induced liver fibrosis.

To further prove the role of the HMGB1 isoforms in liver fibrosis, and circumvent the potential limitation of isoform degradation after injection, we engineered AAV8 vectors. Hmgb1^ΔHepΔMye mice were transduced with AAV8.Tbg.WT.Hmgb1.Gfp to overexpress WT or native HMGB1, which can produce all HMGB1 isoforms, or with AAV8.Tbg.NLS1(2C→2S).Gfp to overexpress Δ[O] HMGB1, which cannot undergo oxidation. Expression of HMGB1 was confirmed by GFP fluorescence in total liver and primary hepatocytes, and ruled out in Kupffer cells isolated from these mice (Fig. S4B top). Likewise, immunofluorescence showed co-localization of HMGB1 with HNF4α (hepatocyte marker), but not with F4/80 (Fig. S4B bottom). HMGB1 morphometry showed similar HMGB1 expression in both groups (Fig. S4C). Two weeks after, mice were injected with CCl₄ for 1 month. Liver injury and fibrosis assessed by H&E staining and IHC for HMGB1, COL1, DES, F4/80 and TNFA (Fig. 3A), the pathology scores (Fig. 3B left), quantification of COL1 by morphometry and western blot (Fig. 3B right, 3C), and ALT activity (Fig. 3D), were significantly reduced in CCl₄-injected Hmgb1^ΔHepΔMye mice transduced with Δ[O] HMGB1, compared to WT HMGB1. RNAseq and IPA revealed decreased proinflammatory and profibrogenic signaling, and improvement of hepatic metabolic function, in CCl₄-injected Hmgb1^ΔHepΔMye mice overexpressing Δ[O] compared to WT HMGB1 (Fig. 3E, Fig. S5). To further confirm the role of [H] and [O] HMGB1 in liver fibrosis progression using a genetic approach, we generated conditional [WT] Hmgb1 (overexpress WT HMGB1 and can produce all isoforms) and Δ[O] Hmgb1 (cannot produce [O] HMGB1) knockin mice in hepatocytes. Alb.Cre, [WT] Hmgb1^Stop.fl/fl, [WT] Hmgb1^{KI Hep}, Δ[O] Hmgb1^Stop.fl/fl and Δ[O] Hmgb1^{KI Hep} mice, were injected CCl₄ for 1 month. Overexpression of HMGB1 in hepatocytes was confirmed by mCherry fluorescence and total HMGB1 by IF (Fig. 3F top four panels). Liver injury and fibrosis assessed by H&E staining and IHC for COL1 (Fig. 3F bottom two panels), the pathology scores and quantification of COL1 by morphometry (Fig. 3F table), further established that reducing hepatocyte production of [O] HMGB1, decreases progression of liver fibrosis. Nonetheless, whether the effects of the HMGB1 isoforms, specifically of [O] HMGB1, were receptor mediated, and how they triggered a fibrogenic response remained elusive.

Fig. 3. Overexpression of WT HMGB1 in hepatocytes promotes, whereas ablation of [O] HMGB1 reduces CCl₄-induced liver fibrosis.

Hmgb1^ΔHepΔMye mice were transduced with AAV8.Tbg.WT.Hmgb1.Gfp (WT HMGB1) or AAV8.Tbg.NLS1(2C→2S).Gfp (Δ[O] HMGB1). Two weeks later, they were injected with CCl₄ for 1 month. (A) Liver injury assessed by H&E staining (necrosis [blue arrows] and inflammation [yellow arrows]), HMGB1 (green arrows), COL1 (perivenular fibrosis [closed pink arrow] and bridging fibrosis [open pink arrows]), DES (green arrows), F4/80 (green arrows), and TNFA (green arrows) IHC. (B) Necrosis, hepatocyte ballooning degeneration, inflammation, and fibrosis scores, and COL1 morphometry. (C) Western blot analysis for COL1. Results are expressed as fold-change of WT HMGB1 assigned a value of 1. (D) Serum ALT activity. (E) Pathway analysis from whole liver RNAseq data using IPA. Red and blue represent positive or negative Z-scores, indicating positive or negative regulation of the pathway, respectively (pink arrows: profibrogenic pathways). (F) Alb.Cre, [WT] Hmgb1^Stop.fl/fl, [WT] Hmgb1^{KI Hep}, Δ[O] Hmgb1^Stop.fl/fl and Δ[O] Hmgb1^{KI Hep} mice, were injected CCl₄ for 1 month. Immunofluorescence of mCherry (red: overexpression of HMGB1) and HMGB1 (green: endogenous plus overexpressed HMGB1), DAPI staining (blue) and the overlay, confirmed the targeting strategy (four top panels). Liver injury assessed by H&E staining (necrosis [blue arrows] and inflammation [yellow arrows]), COL1 (perivenular fibrosis [closed pink arrow] and bridging fibrosis [open pink arrows]) (two bottom panels). Necrosis, hepatocyte ballooning degeneration, inflammation, and fibrosis scores, and COL1 morphometry (table). In (A–D), results are expressed as mean ± SEM; n=3/group; *p<0.05 and **p<0.01 vs. WT HMGB1. In (F), results are expressed as mean ± SEM; n=4/group. *p<0.05 and **p<0.01 vs. Alb.Cre; •p<0.05 and ••p<0.01 vs. [WT] Hmgb1^{KI Hep}.

The HMGB1 isoforms regulate the HSC pro-fibrogenic phenotype and fate.

We previously showed that co-culture with hepatocytes or Kupffer cells from CCl₄-injected WT mice enhanced COL1 production by HSCs, compared to co-culture with cells from MO-injected mice. However, co-culture with hepatocytes or Kupffer cells from CCl₄-injected Hmgb1^ΔHep or Hmgb1^ΔMye mice, respectively, partially prevented this effect [3, 5]. To study how the HMGB1 isoforms affected the pro-fibrogenic response of HSCs, these cells were treated with each isoform for 24 hours. [H] HMGB1 increased total COL1, [O] HMGB1 increased it more, whereas [SO₃] HMGB1 remarkably reduced it, compared to control HSCs (Fig. 4A). The latter raised the hypothesis that HMGB1 sulfonation could be a mechanism to favor fibrosis resolution. To determine if redox-sensitive HMGB1 isoforms affected the invasive potential of HSCs, an important function gained during activation that promotes their pro-fibrogenic phenotype, HSCs were treated with each isoform, in the presence or absence of mitomycin to inhibit mitosis, for 24 hours. [H] HMGB1 significantly increased HSC migration but not [O] HMGB1, and it was not blocked by mitomycin (Fig. 4B). To test whether [O] and [SO₃] HMGB1 competed for the effects on HSCs, cells were treated with a final concentration of 1 nM of HMGB1 at different [O]/[SO₃] ratios (3/0, 2/1, 1/1, 1/2, and 0/3). HSCs expressed less total COL1 and more cleaved CASP3 as [SO₃] increased and [O] HMGB1 decreased (Fig. 4C left), however, the effect was prevented by blocking RAGE (Fig. 4C right). To further proof that the HMGB1 isoforms bind RAGE on HSC, we performed immunofluorescence and colocalization, and proximity ligation assays (PLA), which showed that [H], [O] and [SO₃] HMGB1 bind RAGE, but [O] and [SO3] HMGB1 exhibit greater binding ability (Fig. S6A–B). To identify early events, we performed RNAseq and IPA of HSCs treated with each isoform for 6 hours, finding increased profibrogenic and proinflammatory pathways in HSCs treated with [O] HMGB1, whereas [SO₃] HMGB1 remarkably reduced them (confirmed by proteomics) and induced HSC apoptosis, compared to control HSCs (Fig. S7A–S7D). Consequently, redox-sensitive HMGB1 isoforms play distinct roles on the HSC pro-fibrogenic phenotype and fate, critical events for liver fibrosis progression and resolution.

Fig. 4. The HMGB1 isoforms regulate the HSC pro-fibrogenic phenotype and fate.

HSCs were challenged with 0–1 nM of each endotoxin-free isoform, for 24 hours. (A) Western blot analysis for intra- and extracellular COL1. Results are expressed as fold-change of control assigned a value of 1. (B) Migration of HSCs treated as in in the presence or absence of mitomycin (inhibitor of mitosis). (C) HSCs were treated with a final concentration of 1 nM HMGB1 at different [O]/[SO₃] ratios (3/0, 2/1, 1/1, 1/2, and 0/3) and with or without a RAGE neutralizing antibody. Western blot analysis for intra- and extracellular COL1 and CASP3. In all panels, results are expressed as mean ± SEM; n=3/group; *p<0.05, **p<0.01 and ***p<0.001 vs. control.

[H] and [O] HMGB1 signal through RAGE to upregulate COL1 expression in HSCs.

To identify how redox-sensitive HMGB1 isoforms increased COL1 expression, we evaluated whether it was receptor-mediated. Among the receptors HMGB1 binds to, RAGE and TLR4 are expressed in HSCs, and participate in liver fibrosis [3, 5, 13]. HSCs were isolated from Rage^ΔHSC mice, Tlr4 was silenced in HSCs with small hairpin RNA (shRNA), and cells were challenged with [H] or [O] HMGB1, for 24 hours. Western blot analysis revealed that RAGE (Fig. 5A) but not TLR4 (not shown), was critical for [H] or [O] HMGB1 to stimulate COL1 production in HSCs. The HMGB1 isoforms did not alter MMP1 expression, the metalloproteinase that specifically cleaves COL1 (not shown). Therefore, [H] and [O] HMGB1 upregulate COL1 in HSCs in vitro through RAGE.

Fig. 5. [H] and [O] HMGB1 signal through RAGE to upregulate COL1 expression in HSCs.

(A) Western blot for intra- and extracellular COL1 and RAGE in HSCs from control and Rage^ΔHSC mice, treated with 0–1 nM of endotoxin-free [H] or [O] HMGB1, for 24 hours. Results are expressed as fold-change of control assigned a value of 1. n=3/group; *p<0.05 and **p<0.01 vs. control; •p<0.05, ••p<0.01 and •••p<0.001 vs. WT.Rage ablation in HSCs prevents liver fibrosis in vivo. (B top) Rage^ΔHSC mice were injected i.p. 0–0.1 μg/g of BSA or endotoxin-free [H] or [O] HMGB1, 1 hour before each CCl₄ dose, for 1 month. Co-localization of RAGE (red staining) with DES (HSC marker, green cytoplasmic staining) to validate the targeting strategy. (B bottom) H&E staining showing necrosis (blue arrows) and inflammation (yellow arrows), and IHC for COL1 showing perivenular fibrosis (closed pink arrows) and bridging fibrosis (open pink arrows), in co-treated Rage^ΔHSC mice. (C) Necrosis, hepatocyte ballooning degeneration, inflammation, and fibrosis scores, and COL1 morphometry, and (D) western blot in co-treated Rage^ΔHSC mice. Results are expressed as mean ± SEM; n=6/group; *p<0.05 vs. WT + CCl₄; •p<0.05 vs. control.

To further investigate the role of RAGE on the effects of redox-sensitive HMGB1 isoforms in vivo, Rage^ΔHSC (Fig. 5B top) and Tlr4^ΔHSC (Fig. S8A top) mice, were co-injected [H] or [O] HMGB1, before each CCl₄ dose, for 1 month. Liver injury and fibrosis assessed by H&E staining and COL1 IHC (Fig. 5B bottom), the pathology scores (Fig. 5C left), quantification of COL1 by morphometry (Fig. 5C right) and western blot (Fig. 5D), were significantly reduced in Rage^ΔHSC co-injected either isoform compared to WT mice; however, Tlr4^ΔHSC mice were unprotected from the effects of these two HMGB1 isoforms (Fig. S8A–S8C), compared to mice injected CCl₄ alone. Thus, RAGE was recognized as the main receptor through which these isoforms signal during liver fibrosis progression.

Injection of [SO₃] HMGB1 promotes, while HMGB1 neutralization prevents fibrosis resolution.

Due to the pro-apoptotic effect of [SO₃] HMGB1 in HSCs (Fig. 4C), we investigated whether treating mice with this isoform could promote fibrosis resolution. WT mice injected CCl₄ for 1 month, were allowed to recover for 1 week, while being daily i.p. injected BSA, [SO₃] HMGB1, or an Ab to neutralize [SO₃] HMGB1. H&E staining, HMGB1 and COL1 IHC (Fig. 6A), the pathology scores (Fig. 6B left), HMGB1 morphometry (Fig. S9A), quantification of COL1 by morphometry and western blot (Fig. 6B right, 6C), and ALT activity (Fig. 6D), revealed efficient fibrosis resolution in mice injected [SO₃] HMGB1, compared to mice resolving by themselves. The decrease in ALT with the HMGB1 Ab may reflect liver regeneration, despite persisting fibrosis (Fig. 6D). Mice with HMGB1 neutralization showed similar resolution than those injected BSA. Co-localization of DES⁺TUNEL⁺ cells, quantification of the number of double positive cells over time, and the ratio to total DES⁺ cells, suggested that HSC apoptosis was a key event in these effects, which was confirmed by reduced DES (Fig. 6E). RNAseq and IPA revealed decreased profibrogenic and increased proapoptotic pathways by injecting [SO₃] HMGB1 during fibrosis resolution, whereas injecting HMGB1 Ab prevented them (Fig. 6F, Fig. S9B). Hence, injecting [SO₃] HMGB1 promotes, while neutralizing it prevents fibrosis resolution by inducing HSC apoptosis.

Fig. 6. Injection of [SO₃] HMGB1 promotes, while HMGB1 neutralization prevents fibrosis resolution.

WT mice were injected with MO or CCl₄ for 1 month, to allow onset of peak fibrosis. Upon cessation of CCl₄ injections, mice were allowed to recover from fibrosis while being injected i.p. 0–0.1 μg/g of BSA or endotoxin-free [SO₃] HMGB1, or 0.3 μg/g of a neutralizing Ab to HMGB1, every day, up to 1 week. (A) H&E staining showing necrosis (blue arrows) and inflammation (yellow arrows). IHC for HMGB1 (green arrows), and COL1 showing perivenular fibrosis (closed pink arrows) and bridging fibrosis (open pink arrows). (B) Necrosis, hepatocyte ballooning degeneration, inflammation, and fibrosis scores, and COL1 morphometry. (C) Western blot analysis for COL1. Results are expressed as fold-change of progression (CCl₄) assigned a value of 1. (D) Serum ALT activity. (E top) Co-localization of DES (HSC marker, green cytoplasmic staining) with TUNEL (to detect apoptosis, red nuclear staining), showing loss of DES staining in mice treated with [SO₃] HMGB1 due to HSC apoptosis, and improved resolution. Time-course showing the number of DES⁺TUNEL⁺ cells from peak fibrosis up to 1 week of resolution. (E lower left) Quantification of the number of apoptotic HSCs (DES⁺TUNEL⁺) in 10 fields at 400x, and (E lower right) ratio of number of apoptotic HSCs to total number of HSCs (DES⁺). (F) Pathway analysis from whole liver RNAseq data using IPA. Red and blue represent positive or negative Z-scores, indicating positive or negative regulation of the pathway, respectively (pink arrows: profibrogenic pathways; green arrows: pathways involved in cell cycle). In all panels, results are expressed as mean ± SEM; n=6/group; *p<0.05, **p<0.01 and ***p<0.001 vs. WT + CCl₄; •p<0.05 and ••p<0.01 vs. BSA resolution.

RAGE is involved in the protective effects of [SO₃] HMGB1 during fibrosis resolution.

To further dissect if the effects of [SO₃] HMGB1 on HSC apoptosis during fibrosis resolution, were receptor mediated, Rage^ΔHSC and Tlr4^ΔHSC mice were injected [SO₃] HMGB1 during the resolution phase. H&E staining, HMGB1 and COL1 IHC (Fig. 7A), the pathology scores (Fig. 7B left), HMGB1 morphometry (Fig. S10), quantification of COL1 by morphometry and western blot (Fig. 7B right, 7C), and ALT activity (Fig. 7D), revealed similar fibrosis in Rage^ΔHSC mice injected [SO₃] HMGB1, compared to BSA-injected Rage^ΔHSC mice undergoing fibrosis resolution. Conversely, fibrosis was reduced in Tlr4^ΔHSC mice injected [SO₃] HMGB1, compared to BSA-injected Tlr4^ΔHSC mice undergoing fibrosis resolution (Fig. S11). These results suggested that the pro-apoptotic effects of [SO₃] HMGB1 on HSCs were mediated by RAGE. Co-localization of DES⁺TUNEL⁺, quantification of the number of double positive cells, and the ratio to total DES⁺ cells validated that HSC apoptosis was a key event in this effect, which was confirmed by reduced DES (Fig. 7E, Fig. S11F). Therefore, [SO₃] HMGB1 promotes fibrosis resolution signaling through RAGE in HSCs to induce apoptosis.

Fig. 7. RAGE is involved in the protective effects of [SO₃] HMGB1 during fibrosis resolution.

Rage^ΔHSC mice were injected MO or CCl₄ for 1 month to reach peak fibrosis. Following cessation of CCl₄ injections, mice were allowed to recover from fibrosis, while being injected i.p. 0–0.1 BSA or μg/g of endotoxin-free [SO₃] HMGB1, daily, for 1 week. (A) H&E staining showing necrosis (blue arrows) and inflammation (yellow arrows). IHC for HMGB1 (green arrows), and COL1 showing perivenular fibrosis (closed pink arrows) and bridging fibrosis (open pink arrows). (B) Necrosis, hepatocyte ballooning degeneration, inflammation, and fibrosis scores, and COL1 morphometry. (C) Western blot analysis for COL1. Results are expressed as fold-change of progression (CCl₄) assigned a value of 1. (D) Serum ALT activity. (E top)Co-localization of DES (marker of HSC, green cytoplasmic staining) with TUNEL (to detect apoptosis, red nuclear staining), demonstrating less staining in Rage^ΔHSC mice co-treated with [SO₃] HMGB1, compared to values for WT mice shown in Fig. 6E. (E bottom left) Quantitative analysis of the number of apoptotic HSC (DES⁺TUNEL⁺) in 10 fields at 400x, and (E bottom right) ratio of number of apoptotic HSC to total number of HSC (DES⁺). In all panels, results are expressed as mean ± SEM; n=4/group; *p<0.05, **p<0.01 and ***p<0.001 vs. WT + CCl₄; •p<0.05, ••p<0.01 and •••p<0.001 vs. Rage^ΔHSC + CCl₄.

Gene expression signatures induced by redox-sensitive HMGB1 isoforms in mice classify fibrotic patients according to fibrosis and inflammation scores.

To determine the clinical relevance of our findings, the expression of genes modified by redox-sensitive HMGB1 isoforms in mice was analyzed in a cohort of 124 Hepatitis B patients [14]. Following hierarchical clustering of genes differentially expressed in Hmgb1^ΔHepΔMye mice overexpressing Δ[O] HMGB1 compared to WT HMGB1 in fibrosis progression, a cluster of patients with high fibrosis and inflammation scores was identified (Fig. 8A). These patients highly expressed genes decreased in Hmgb1^ΔHepΔMye mice overexpressing Δ[O] HMGB1 during fibrosis progression. Moreover, genes decreased in WT mice injected [SO₃] HMGB1, compared to BSA during fibrosis resolution, identified a cluster of patients with high fibrosis and inflammation scores (Fig. 8B). Overall, these results highlight that redox-sensitive HMGB1 isoforms found in this study could also play a role in fibrosis progression and resolution in humans.

Fig. 8. Gene expression signatures induced by redox-sensitive HMGB1 isoforms in mice classify fibrotic patients based on fibrosis and inflammation scores.

mRNA expression from 124 Hepatitis B fibrotic patients, was analyzed by hierarchical clustering with the Ward’s method and the Pearson correlation coefficient, to generate dendrograms. Heatmap showing the expression profile of paralog genes, to those found differentially expressed in Hmgb1^ΔHepΔMye mice overexpressing Δ[O] vs WT HMGB1 during fibrosis progression (A left). Heatmap showing the expression profile of paralog genes, to those found differentially expressed in WT mice injected [SO₃] HMGB1 vs BSA during fibrosis resolution (B left). Low and high gene expression genes (median as cutoff) are represented in green and red, respectively. After clustering patients into two groups, the difference in the Scheuer fibrosis and inflammation scores between groups was determined by the chi-square test (A and B right).

DISCUSSION

To date, management, prevention, and reversal of liver fibrosis is an unmet clinical need. To accelerate preclinical development of potent anti-fibrotic therapies, reliable single target validation is vital. Moreover, a key challenge limiting progress in testing anti-fibrotic drugs is lack of sufficient noninvasive endpoints correlating well with clinical outcome. A key finding from this study is the identification of signature redox-sensitive HMGB1 isoforms ([H], [O], [SO₃]) present in liver from mice with liver fibrosis progression and resolution.

Activation of thioredoxins (TXN) plays a role in HMGB1 oxidation by maintaining the activity of peroxiredoxins (PRDX) [15]. Indeed, Txn2, Txnrd1, Txnrd2 and Txndc12 mRNAs increased, whereas Prdx1 and Prdx2 mRNAs, encoding proteins that induce disulfide bond formation in HMGB1 [16], remained similar in WT mice injected CCl₄ or MO (not shown). Increased reactive oxygen and nitrogen species, generated during liver injury, lead to protein sulfonation. Whether this modification increases protein stability, favoring accumulation, remains unknown. The appearance of [SO₃] HMGB1 may result from prolonged endoplasmic reticulum stress, or from induction of sulfotransferases due to liver regeneration [17].

This study proves that hepatocytes are the main source of all redox-sensitive HMGB1 isoforms whereas myeloid cells mostly produce [H] HMGB1. While significant efforts focused on understanding the role of HSCs in the pathogenesis of liver fibrosis, the contribution of neighboring liver cells to HSC activation, and reversal or clearance, has received less attention. Here, we provide new insight on the interplay between liver cells that drives the pathogenesis of liver fibrosis. Specifically, we dissect how redox-sensitive dynamic changes in hepatocyte-derived HMGB1 isoforms signal through RAGE-dependent mechanisms to drive the HSC pro-fibrogenic phenotype and fate, therefore, contributing to liver fibrosis progression and resolution.

First, we show that [H] HMGB1 stimulates HSC migration, a critical feature gained by HSCs that supports fibrogenesis. Second, we prove that administration of [O] HMGB1 enhances liver fibrosis more than [H] HMGB1, and that overexpression of WT or native HMGB1 in hepatocytes promotes, whereas ablation of [O] HMGB1 in hepatocytes reduces liver fibrosis. Furthermore, [O] HMGB1 stimulates macrophage proliferation and TNFA expression, both pro-inflammatory events involved in fibrosis progression. Third, we establish that administration of [SO₃] HMGB1 aggravates, while HMGB1 neutralization prevents fibrosis resolution, thus this isoform is critical for fibrosis reversal. The appearance of [SO₃] HMGB1 is not just a consequence of resolution or a mere bystander, but a causative factor that triggers HSC apoptosis with the ensuing COL1 decrease, and eventual reduction in scar formation.

Although we identified that HMGB1 undergoes redox-sensitive dynamic PTMs, critical for fibrosis progression and resolution, yet the receptor involved remained elusive. Since HMGB1 conveys signals to the host through cell surface receptors [3, 5, 13], it was likely that the identified HMGB1 isoforms acted through receptor binding along with downstream pro-fibrogenic events. This study unveils that these isoforms bind and signal through RAGE in HSCs to stimulate migration ([H] HMGB1) and upregulate COL1 production ([H] and [O] HMGB1), during fibrosis progression. It also reveals that [SO₃] HMGB1 signals through RAGE to trigger CASP3-dependent apoptosis, during fibrosis resolution. This was established in vitro and by co-injection of [H] and [O] HMGB1 during fibrosis progression, and by injection of [SO₃] HMGB1 during fibrosis resolution, to Rage^ΔHSC, Tlr4^ΔHSC and WT mice.

Overall, our study highlights a major role for a sterile damage-associated molecular pattern, as a redox biosensor undergoing changes in oxidation state throughout the course of the disease in mice. Further, we found correlation with progression of liver fibrosis in humans. Conceivably, a rise in [H] HMGB1 could flag ‘patient at risk’, the presence of [O] HMGB1 could suggest ‘disease in progress or active scarring’, while the appearance of [SO₃] HMGB1 could point at ‘resolution under way’. The latter could be used as a readout for response to pharmacological intervention with anti-fibrotic agents. From the clinical standpoint, the prospect of identifying and/or predicting active versus regressive fibrosis, by monitoring the appearance of redox-sensitive HMGB1 isoforms over time, is clinically significant.

Dynamic changes in the isoforms could be used as readouts of clinical course of the disease, for patient stratification, and treatment outcome, and to identify individuals at greater risk of fibrosis progression. It is likely that initial blockade of [H] HMGB1 will reduce HSC migration. As injury progresses, oxidation will be rather quick, thus neutralization of [O] HMGB1 and/or RAGE, or co-treatment with an antioxidant, will block inflammation, macrophage infiltration, TNFA production, and COL1 increase. Lastly, administration of [SO₃] HMGB1 or activation of RAGE will favor fibrosis resolution by inducing HSC apoptosis therefore lessening scarring (Fig. S12).

Overall, these results provide compelling evidence to propose a major role for redox-sensitive dynamic changes in HMGB1 isoforms in liver fibrosis progression and resolution. HMGB1, unlike other pro-fibrogenic mediators, provides a much wider time-frame for clinical intervention due to its remarkable cellular mobility and prolonged half-life [18], making it easy to be detected in liver, and ultimately causing longer receptor activation. Hence, these redox-sensitive HMGB1 isoforms and RAGE are attractive targets to prevent fibrosis progression and promote resolution.

IMPACT AND IMPLICATIONS.

Since oxidant stress plays a major role in liver fibrosis and induces post-translational modifications of proteins, we hypothesized that redox-sensitive HMGB1 isoforms contribute to liver fibrosis progression and resolution. This study is significant because a rise in [H] HMGB1 could flag ‘patient at risk’, the presence of [O] HMGB1 could suggest ‘disease in progress or active scarring’, while the appearance of [SO₃] HMGB1 could point at ‘resolution under way’. The latter could be used as a readout for response to pharmacological intervention with anti-fibrotic agents.

HIGHLIGHTS.

• We identified the post-translational modifications of HMGB1 during liver fibrosis progression ([O] and [SO₃] HMGB1) and resolution ([SO₃] HMGB1).

• The pro-fibrogenic effect of [O] HMGB1 occurs through RAGE signaling in hepatic stellate cells.

• [SO₃] HMGB1 accelerates fibrosis resolution due to RAGE-dependent stimulation of hepatic stellate cell apoptosis.

• Gene signatures, activated by redox-sensitive HMGB1 isoforms in mice, classify fibrotic patients according to fibrosis and inflammation scores.

ABBREVIATIONS

[H] HMGB1

native or fully-reduced HMGB1

[O] HMGB1

disulfide HMGB1

[SO₃] HMGB1

sulfonated HMGB1

[WT] Hmgb1 ^{KI Hep}

conditional WT Hmgb1 knockin mice in hepatocytes

[WT] Hmgb1 ^Stop.fl/fl

WT Hmgb1 Stop floxed mice

AAV8

adeno-associated virus serotype 8

Ab(s)

antibody(ies)

Alb

albumin

ALT

alanine aminotransferase

CCl₄

carbon tetrachloride

Cys

cysteine

ESI-LC-MS

electrospray ionization-liquid chromatography-mass spectrometry

GFP

green fluorescent protein

GSEA

gene set enrichment analysis

H&E

hematoxylin and eosin

Hep(s)

hepatocyte(s)

HMGB1

high-mobility group box-1

Hmgb1 ^fl/fl

Hmgb1 floxed mice

Hmgb1 ^ΔHep

conditional Hmgb1 knockout mice in hepatocytes

Hmgb1 ^ΔHepΔMye

conditional Hmgb1 knockout mice in hepatocytes and myeloid cells

Hmgb1 ^ΔMye

conditional Hmgb1 knockout mice in myeloid cells

HSC(s)

hepatic stellate cell(s)

IHC

immunohistochemistry

IPA

ingenuity pathway analysis

Lrat

lecithin retinol acyltransferase

Lyz2

lysozyme-2

MO

mineral oil

NLS

nuclear localization signal

PLA(s)

proximity ligation assay(s)

PTM(s)

post-translational modification(s)

RAGE

receptor for advanced-glycation end-products

Rage ^fl/fl

Rage floxed mice

Rage ^ΔHSC

conditional Rage knockout mice in hepatic stellate cells

RNAseq

RNA sequencing

SAMe

S-adenosylmethionine

Ser

serine

shRNA

small hairpin RNA

Tbg

thyroxine-binding globulin

TLR4

toll-like receptor-4

Tlr4 ^fl/fl

Tlr4 floxed mice

Tlr4 ^ΔHSC

conditional Tlr4 knockout mice in hepatic stellate cells

TXN

thioredoxins

WT

wild-type

Δ[O] HMGB1

oxidation-incompetent HMGB1 mutant

Δ[O] Hmgb1 ^{KI Hep}

conditional Δ[O] Hmgb1 knockin mice in hepatocytes

Δ[O] Hmgb1 ^Stop.fl/fl

Δ[O] Hmgb1 Stop floxed mice.

DATA AVAILABILITY STATEMENT

All data sets (GSE233751) are available in the NCBI Gene Expression Omnibus database (www.ncbi.nlm.nih.gov/geo).