FGF21 ameliorates septic liver injury by restraining proinflammatory macrophages activation through the autophagy/HIF-1α axis

Graphical abstract

Schematic showing that FGF21 alleviates septic liver injury by inhibiting proinflammatory activation of macrophages by promoting autophagic degradation of HIF-1α.

Highlights

• •FGF21 expression in hepatocytes and macrophages is upregulated during sepsis.
• •FGF21 inhibits proinflammatory macrophage activation by resuming autophagy flux to ameliorate septic liver injury.
• •FGF21 accelerates autolysosome degradation via mTOR signaling, while it has no effect on autophagosome formation induced by LPS via ERK and JNK signaling pathways.
• •FGF21 promotes p62-dependent autophagic degradation of HIF-1α in macrophages by enhancing the interaction between p62 and HIF-1α.

Abstract

Introduction

Sepsis, a systemic immune syndrome caused by severe trauma or infection, poses a substantial threat to the health of patients worldwide. The progression of sepsis is heavily influenced by septic liver injury, which is triggered by infection and cytokine storms, and has a significant impact on the tolerance and prognosis of septic patients. The objective of our study is to elucidate the biological role and molecular mechanism of fibroblast growth factor 21 (FGF21) in the process of sepsis.

Objectives

This study was undertaken in an attempt to elucidate the function and molecular mechanism of FGF21 in therapy of sepsis.

Methods

Serum concentrations of FGF21 were measured in sepsis patients and septic mice. Liver injury was compared between mice FGF21 knockout (KO) mice and wildtype (WT) mice. To assess the therapeutic potential, recombinant human FGF21 was administered to septic mice. Furthermore, the molecular mechanism of FGF21 was investigated in mice with myeloid-cell specific HIF-1α overexpression mice (LyzM-Cre^DIO-HIF-1α) and myeloid-cell specific Atg7 knockout mice (Atg7^△mye).

Results

Serum level of FGF21 was significantly increased in sepsis patients and septic mice. Through the use of recombinant human FGF21 (rhFGF21) and FGF21 KO mice, we found that FGF21 mitigated septic liver injury by inhibiting the initiation and propagation of inflammation. Treatment with rhFGF21 effectively suppressed the activation of proinflammatory macrophages by promoting macroautophagy/autophagy degradation of hypoxia-inducible factor-1α (HIF-1α). Importantly, the therapeutic effect of rhFGF21 against septic liver injury was nullified in LyzM-Cre^DIO-HIF-1α mice and Atg7^△mye mice.

Conclusions

Our findings demonstrate that FGF21 considerably suppresses inflammation upon septic liver injury through the autophagy/ HIF-1α axis.

Introduction

Sepsis, a life-threatening disease induced by severe trauma and infection, frequently occurs in intensive care units (ICU) and its mortality rate can reach nearly 25 %, which places a huge burden on global medical and healthcare systems [1], [2]. In clinical settings, anti-bacterial and anti-inflammatory strategies are commonly used to deal with sepsis. Combinatorial use of antibiotics and hormones, liquid resuscitation, vasopressin administration, mechanical ventilation and even replacement of failed organs can effectively reduce the mortality of sepsis patients [3], [4]. Although the development of medical technology on today provides the possibility for interventional treatment of sepsis, abuse of antibiotics and hormones and mechanical injury induced by respiratory intervention cause abundant complications. Therefore, new therapies for sepsis using safer and more effective drugs are still need to be explored.

During the progression of sepsis, the liver exposed to circulating pathogens, endotoxins, and inflammatory cytokines because of its anatomical characteristics and special circulation structure, thus is one of the most vulnerable organs to sepsis [5]. Upon activation of innate immune defense constructed by endothelial cells, neutrophils, Hepatic macrophages (Kupffer cells and monocyte-derived-macrophages (MDMs), the liver switches from metabolic function to immunogenic responses and mediates the production and releases of acute phase proteins to adapt to the septic environment [5], [6], [7], [8]. During immune defense and clearance of invasive pathogens, macrophage-produced excessive inflammatory cytokines (inflammatory storm) cause indiscriminate damage to normal tissues, which result in organ failure and death [9], [10]. Therefore, inhibition of the macrophage mediated excessive inflammatory response can effectively maintain normal liver function during sepsis, reduce the mortality of sepsis patients, and improve their prognosis [11].

As an endogenous metabolic regulator mainly expressed in the liver, fibroblast growth factor 21 (FGF21) regulates some metabolic disorders (diabetes, obesity and atherosclerosis) and elicits protective effects against some acute diseases (acetaminophen induced acute liver injury, myocardial infarction and stroke) [12], [13], [14], [15]. A previous study showed that FGF21 deficient mice (FGF21 KO mice) had higher mortality after lipopolysaccharide (LPS) injection, indicating that FGF21 elicits a protective effect on sepsis [16]. Although the underlying mechanism is unknown, studies show that FGF21 can inhibit the inflammatory activation of macrophages [17], [18]. Thereby, we explored whether FGF21 ameliorates septic liver injury by regulating proinflammatory macrophage activation and the detailed underlying molecular mechanism.

In this study, we found that FGF21 was upregulated both in hepatocytes and macrophages in the liver of sepsis mice. By using CLP (a mouse model of sepsis), we demonstrated that FGF21 promotes the autophagic degradation of HIF-1α to suppress the proinflammatory activation of macrophages, which ameliorates inflammatory liver injury and reduces sepsis mortality. Moreover, we demonstrated that PF-05231023 (a long-acting analog of FGF21) elicits similar protective effect in septic liver injury and reduces the mortality of mice from sepsis.

Materials and methods

Ethics statement

Before collection the sample from healthy volunteers and sepsis patients, we provided the human patient consent forms to healthy volunteers and sepsis patients. And patients’ s data and pathologic evaluations were approved by the Ethics Committee of the Second Affiliated Hospital and Yuying Children's Hospital of Wenzhou Medical University (Approval ID: 2021-K-106–01). All animals in our study received humane care according to the criteria outlined in the “Guide for the Care and Use of Laboratory Animals”. Animal protocols were approved by the Ethics Committee of Wenzhou Medical University.

Mice cohorts and treatment

C57BL/6 male mice (8–10 weeks, 20–25 g) in our study were purchased from Model Animal Research Center of Nanjing University. C57BL/6 male mice were divided into the following groups: (1) sham-operated group; (2) CLP-operated and 0.9 % saline-injected intraperitoneally group; (3) CLP-operated and rhFGF21(2 mg/kg)-injected intraperitoneally group. The rhFGF21 used in our study was produced by Wenzhou Medical University Gene Engineering Laboratory [19]. To evaluate the therapeutic potential of PF-05231023 (Abmole, M10048), C57BL/6 male mice were divided into the following groups: (1) sham-operated group; (2) CLP-operated and 0.9 % saline-injected intraperitoneally group; (3) CLP-operated and PF-05231023 (10 mg/kg)-injected intraperitoneally group.

Male FGF21 KO mice (8–10 weeks, 20–25 g) and WT control were provided by Dr. Steve Kliewer. To genetically upregulate FGF21 in liver of mice, we injected AAV2/8-Fgf21 (5 × 10¹² vg/ml) intravenously to four-week-old Lyzm-Cre mice. Steadily overexpressed for three weeks, AAV2/8-Fgf21 −injected mice and AAV2/8-vector-injected mice were prepared to subsequent experiments.

The LyzM-Cre mice were obtained from Jackson Laboratories (004781). To generate LyzM-Cre^DIO-HIF-1α mice, we injected AAV2/8-CMV-DIO-HIF-1α-P2A-EGFP-tWPA (5 × 10¹² vg/ml) intravenously to four-week-old Lyzm-Cre mice. Steadily overexpressed for three weeks, AAV2/8-CMV-DIO-HIF-1α-P2A-EGFP-tWPA-injected mice and AAV2/8-CMV-DIO-EGFP-tWPA-injected mice were prepared to subsequent experiments. These adeno-associated virus were constructed by OBIO technology (Shanghai), and the clone serial number are H19230 (AAV2/8-CMV-DIO-EGFP-tWPA) and H21562 (AAV2/8-CMV-DIO-HIF-1α-P2A-EGFP-tWPA). And the reference sequence of AAV2/8-CMV-DIO-HIF-1α-P2A-EGFP-tWPA is NM_010431.2 (GenBank ID).

Atg7 flox/flox mice was a kind gift from Dr. Yuqiang Ding. And we crossed Atg7 flox/flox mice with LyzM-Cre mice to generate Atg7^△mye mice. For experimental purposes, Atg7^△mye mice and Atg7 flox/flox littermates were used.

Cell culture

To generate BMDMs, the tibia and femur of male mice were collected. The bone marrow cells were lavaged with RPMI-1640 medium (Sigma Aldrich, R6504) by 10 ml syringe coupled with 25G needle, and passed through a 70 μm strainer, washed, centrifuged. Then the cells were cultivated with complete RPMI-1640 medium (10 % Fetal Bovine Serum (FBS) (Gibco, 10437–028) and 100 units per ml penicillin and 100 μg/ml streptomycin (Invitrogen, Gibco^TM, 15140)). To purified isolated cells, the adherent cells were discarded and the supernatant was collected to centrifuge after 4 h. The isolated cells were cultivated with complete DMEM media (Sigma Aldrich, D1152) (10 % FBS and 20 % of L929 supernatant) for 7 days. To differentiated isolated cells into macrophages, additional media was added to Petri dishes on day 3, day 5 and after 7 days [20]. Kupffer cells from liver of mice were isolated according to the protocol as previously [21].

To isolate primary hepatocytes, we performed liver perfusion with ethylene glycol tetra acetic acid (EGTA) buffer followed by collagenase IV (Sigma Aldrich, C4-BIOC) digestion in C57BL/6 male mice. Then, we dissected the digested liver in a sterile petri dish containing cold DMEM media. After passed through a 70 μm strainer, and centrifuged (5 min, 4 °C), the collected cell was cultivated in complete RPMI-1640 medium [22].

Raw264.7 cell line was purchase from ATCC (ATCC, TIB-71^TM), and cultivated in complete RPMI-1640 medium. LX-2 cell line was purchase from Procell, and cultivated in complete DMEM medium.

LPS (1 μg/ml; Sigma Aldrich, L6386) was used to stimulate BMDMs, primary hepatocytes and LX-2 in different time point. The concentration of rhFGF21 cotreated with LPS in this study was 200 ng/ml. And the concentration of PF-05231023 cotreated with LPS in this study was 100 ng/ml. And in signaling pathway analysis, each agonist and antagonist: cobalt chloride (100 nM; Sigma Aldrich, 255599), U0126 (20 mM; Sigma Aldrich, 19–147), SP600125 (10 mM; Sigma Aldrich, 420119), SB203580 (10 mM; Sigma Aldrich, S8307), BGJ398 (100 nM; Selleck, S2183), Compound C (1 μM; Selleck, S7840) or MHY1485 (10 mM; Selleck, S7811) was pretreated for 12 h before LPS administration. In analysis of HIF-1α degradation, cycloheximide (100 nM, S7418) was pretreated for 12 h before LPS administration, Bafilomycin A1 (100 nM; Sigma Aldrich, 131793) or MG132 (10 μM, Selleck, S2619) was added for last 6 h.

Cecal ligation and puncture model

In brief, C57BL/6 male mice (8–10 weeks) were anesthetized with an intraperitoneal (i.p.) injection of pentobarbital sodium (50 mg/kg). Sham-operated mice underwent identical procedures, except for the CLP. In short-term survival CLP model, the mice were performed high-grade sepsis which caecum was ligated 75 % and the ligated cecal stump was then perforated by one “through and through” punctures (18-gauge needle). The serum sample of high-grade sepsis mice was harvested after 6 h for further experiment. In mild-grade sepsis model, the caecum of mice was ligated 50 % and the ligated cecal stump was then perforated by one “through and through” punctures (22-gauge needle) [23], [24]. 0.9 % saline, rhFGF21 (2 mg/kg) or PF-05231023 (10 mg/kg) was administrated to mice after CLP operation. The serum sample and tissue sample (liver, lung, kidney, heart, iWAT, pancreas and muscle) was harvested after 12 h for further experiment. After collecting the blood sample, the mice underwent perfusion through the left ventricle with 10 ml of 0.9 % saline for 2 min. After perfusion, the lungs, heart, liver, pancreas, kidneys, muscles, and iWAT were sequentially extracted from the mice. Following this, the lungs were positioned in a 10 cm diameter culture dish filled with 0.9 % saline, and images of the lungs' appearance were taken. The remaining tissues were rinsed with physiological saline to eliminate surface blood stains, and then fixed or stored at low temperatures as per the experimental needs.

Detection of GFP-LC3B and autophagic flux

Raw264.7 cell was transfected Lenti-GFP-LC3B (Obio Technology, H2799) for GFP-LC3B detection, and Lenti-mCherry-GFP-LC3B (Obio Technology, H6687) for autophagic flux detection. After 48 h, each experimental treatment was performed to transfected Raw264.7 cell. Leica TCS SP8 Confocal microscope (Leica, Wetzlar, Germany) was used to observe the phenomenon in each group.

ELISA analysis of biochemical parameters

The Human FGF21 (Shanghai MLBIO Biotechnology Co. ltd, ml058174-1), Mouse FGF21 (Boster Biological Technology Co. ltd, EK1379), Mouse-IL-1β (Proteintech, KE10003), Mouse-IL-6 (Proteintech, KE10007), Mouse-TNF-α (Proteintech, KE10002), GPT/ALT (Nanjing Jiancheng Bioengineering Institute, C009-2–1), GOT/AST (Nanjing Jiancheng Bioengineering Institute, C010-2–1)), BUN (Nanjing Jiancheng Bioengineering Institute, C013-2–1) and creatinine (Nanjing Jiancheng Bioengineering Institute, C011-2–1) were analyzed base on the manufacturer’s protocol.

Transmission electron microscopy

Conventionally, BMDMs were fixed (arnofsky fixative), postfixed (1 % osmium tetroxide), dehydrated (acetone), and embedded (Spar resin). Using JEOL 1200 electron microscope (JEOL, Tokyo, Japan) to capture the images from sections of BMDMs (98-nm thickness).

Histological analysis of liver, lungs and kidney section

Mouse livers, lungs and kidney were sectioned at 5 μm. To analyze the proinflammatory cell infiltration in liver, sections from above organs were stained by Hematoxylin-Eosin Kit (Solarbio, G1120). Photographs were taken by PA53 BIO microscope (Motic, PA53).

Immunoblotting analysis

In immunoblotting analysis, 30 μg protein from each sample was resolved by SDS-PAGE on Tris-glycine gels, and transfected to polyvinylidene fluoride membrane. After blocked by 5 % bovine serum albumin (BSA) (sigma, B2064), membranes were incubated with primary antibodies (overnight, 4 °C). And after washed 3 times by TBST, membranes were incubated with either HRP-goat-anti-mouse (Abcam, ab6789) or HRP-goat-anti-rabbit (Abcam, ab6721) secondary antibodies (2 h, room temperature). Primary antibodies included: FGF21 (Abcam, ab171941), iNOS (Cell Signaling Technology, 13120), HIF-1α (Affinity Biosciences, BF8002), LC3 (Cell Signaling Technology, 12741), p62 (Abcam, ab109012), phospho-mTOR (Cell Signaling Technology, 5536), mTOR (Cell Signaling Technology, 2983), phospho-S6 (Cell Signaling Technology, 81736), S6 (Cell Signaling Technology, 2317), phospho-ERK1/2 (Cell Signaling Technology, 4370), ERK1/2 (Cell Signaling Technology, 4695), phosphor-JNK (Cell Signaling Technology, 9255), JNK (Cell Signaling Technology, 9252), TFEB (Cell Signaling Technology, 37785), p-TFEB ((Affinity Biosciences, AF3845), GAPDH (Abcam, ab9485), ULK1 (Cell Signaling Technology, 8054), phosphor-p62 (Affinity Biosciences, DF2985), phosphor-ULK1 (Affinity Biosciences, AF4387) and Lamin B (Cell Signaling Technology, 13435). And the expression of β-actin (Cell Signaling Technology, 3700) was used as a loading control. The protein in each membrane was visualized (pierce ECL plus western blotting substrate (Thermo Scientific, 32132)), analyzed (ChemiDoc MP device (Bio-Rad, Hercules, CA, USA)), and quantified (ImageQuant 5.2 software (Molecular Dynamics, Sunnyvale, CA)).

Immunoprecipitation

500 μg protein of BMDMs were incubated with primary antibody (overnight, 4 °C). Add immobilized antibody beads (Millipore, LSKMAGAG10) to cell lysates for 4 h on next day. Washed beads with TBS for 3 times, and eluted (glycine-HCl (0.1 M, pH 3.5)), then subjected to immunoblotting analysis. Immobilized antibody included: p62 (Santa Cruz Biotechnology, sc-48402) and HIF-1α (Santa Cruz Biotechnology, sc-13515).

Immunohistochemistry staining

After treatment of antigen retrieval (5 min, 96 °C), the liver sections (5 μm) were permeabilized (0.5 % Triton X-100, 30 min, room temperature), blocked (5 % BSA, 30 min, room temperature) and incubated with primary antibodies (overnight, 4 °C). The primary antibodies included: FGF21 (Abcam, ab171941). Then the liver sections were incubated with a biotinylated secondary antibody (2 h, room temperature). Finally, using Diaminobenzidine (DAB) Histochemistry Kit and hematoxylin to stain liver section slides. Photographs were taken by PA53 BIO microscope (Motic, PA53).

Immunofluorescence staining

Before incubated with secondary antibody, the process of immunofluorescence staining is consistent with immunohistochemistry staining. And the primary antibodies included: MPO (Abcam, ab208670), F4/80 (Cell Signaling Technology, 30325), HIF-1α (Santa Cruz Biotechnology, sc-13515), iNOS (Santa Cruz Biotechnology, sc-7271), LC3 (Santa Cruz Biotechnology, sc-376404), p62 (Santa Cruz Biotechnology, sc-48402) and albumin (Proteintech, 66051–1-Ig). After washing 3 times by PBS, liver sections were incubated with secondary antibody (2 h, room temperature). The secondary antibodies included: Goat Anti-Rabbit IgG H&L (Abcam, ab150077), Goat Anti-Mouse IgG H&L (Abcam, ab150113), Goat Anti-Rabbit IgG H&L (Abcam, ab150083) and Goat Anti-Mouse IgG H&L (Abcam, ab150115). DAPI was used to stain the nucleus.

In immunofluorescence staining of cells, BMDMs in each group were fixed (4 % formaldehyde, 15 min, room temperature), permeabilized (0.5 % Triton X-100, 30 min, room temperature), blocked (5 % BSA, 30 min, room temperature), and incubated with primary antibodies (overnight, 4 °C). The primary antibodies included: HIF-1α (Santa Cruz Biotechnology, sc-13515), Lamp1 (Abcam, ab62562) and p62 (Santa Cruz Biotechnology, sc-48402). After washing 3 times by PBS, BMDMs were incubated with secondary antibody (2 h, room temperature). The secondary antibodies used in immunofluorescence staining of BMDMs were consistent with immunofluorescence staining of liver sections. DAPI was used to stain the nucleus. The phenomenon in liver sections and cell sections of each group was analyzed (Leica TCS SP8 Confocal microscope (Leica, Wentzler, Germany)), and quantified (ImageQuant 5.2 software (Molecular Dynamics, Sunnyvale, CA)).

In situ hybridization

The RNA scope Probe of Mm-Fgf21 was purchase from Advanced cell Diagnostia (ACD, 460931). On the basis of the manufacturer’ s protocol, we performed in situ hybridization of Mm-Fgf21 in liver section of mice.

Lysosomal function assay of BMDMs

Before the lysosomal function assay, BMDMs under each experimental condition were washed twice with PBS. Then, BMDMs were incubated with pHLys Red (Dojindo, L265) (30 min, 37 °C). After removing the supernatant, BMDMs were washed twice with PBS before adding complete culture medium. Finally, BMDMs under each experimental condition were observed using a ZEISS LSM 980 with Airyscan2 Confocal microscope (ZEISS, Germany).

Measurement of bacterial counts

Before harvest the tissue of mice, 1 ml PBS was used to lavage the peritoneal cavity of mice in each group under sterile conditions. After collection, we 100-fold diluted the samples of blood and peritoneal lavage fluid (PLF) in sterile 0.9 % saline, and the dilution of blood and PLF were cultured on tryptic soy agar pour plates (BD, BA-256665.02) (48 h, 37 °C). The colony counts were analyzed as described previously [25].

Measurement of cell number in BALF

BALF was collected as described previously [26]. The collect BALF was centrifuged at 2000 g to collect cell for further counting. And the supernatant was used to determine total protein concentration by Pierce BCA Protein Assay Reagent (Thermo Fisher Scientific, 23228).

RNA interference in BMDMs

For RNA interference, the control scramble small interfering RNA or Small interfering RNA of fgf21 (Santa Cruz Biotechnology, SC-39485) or sqstm1 (Santa Cruz Biotechnology, SC-29828) was mix in Opti-MEM (Gibco, 51985034) (20 min, 37 °C). Then, we added the mixture to BMDMs respectively which cultured in Opti-MEM for 8–12 h. Then, the culture medium was changed for another 48 h, followed by subsequent experiments.

TUNEL staining in liver section

Before incubated with TUNEL reaction mixture, the process of TUNEL staining in Liver section is consistent with immunohistochemistry staining. The liver sections were incubated with a TUNEL reaction mixture (1 h, 37 °C) in a humidified atmosphere in the dark. After washed 3 times with PBS, DAPI was used to stain the nucleus. The phenomenon of TUNEL stain in each group was analyzed (Leica TCS SP8 Confocal microscope (Leica, Wentzler, Germany)), and quantified (ImageQuant 5.2 software (Molecular Dynamics, Sunnyvale, CA)).

Statistical analysis

Results are expressed as means ± SEM. Statistical differences were assessed with the unpaired 2-tailed Student t test for two experimental groups and one way ANOVA test for multiple groups. And long-rank test was applied to survival rate experiment. P < 0.05 is considered to be statistically significant. All statistical analyses were done using GraphPad Prism 8 software.

Results

Expression of FGF21 is upregulated in hepatocytes and macrophages in liver of sepsis mice

Analyses of serum samples revealed that the level of FGF21 was significantly increased in serum of patients and mice with sepsis (Fig. 1(A, B)). According to its extensive expression throughout the body, we analyzed the protein level of FGF21 in the liver, heart, inguinal white adipose tissue (iWAT), pancreas and muscle of sepsis mice [27], [28]. In contrast with the findings of a previous study, the protein level of FGF21 was increased in liver but decreased in the heart, iWAT, pancreas and muscle of CLP-operated mice (Fig. 1C). Immunohistochemical analysis of liver sections also indicated that FGF21 was significantly upregulated in CLP-operated mice (Fig. S1A). In addition to upregulating transcription of fgf21 in the liver, CLP promoted transcription of fgfr1 and repressed transcription of β-klotho (a coreceptor of FGF21) (Fig. S1B).

Fig. 1.

The expression of FGF21 in macrophages and hepatocytes is upregulated in liver of septic mice. (A) The serum level of FGF21 in healthy volunteers (n = 5) and sepsis patients (n = 9). (B) The serum level of FGF21 in mice with CLP operation (n = 4). (C) Relative protein level of FGF21 analyzed by western blotting in different organs (liver, heart, iWAT, pancreas and muscle) from mice with CLP operation (n = 4). (D) Representative immunofluorescence of in situ hybridization of FGF21 in liver from mice with CLP operation. DAPI (blue), FGF21(red), and Albumin (green). Scale bars: 50 μm. (E) Representative immunofluorescence of in situ hybridization of FGF21 in liver from mice with CLP operation. DAPI (blue), FGF21(red), and F4/80 (green). White arrows indicate FGF21-expressed macrophages. Scale bars: 50 μm. (F) The expression of FGF21 in primary hepatocytes after stimulated with LPS for various periods of time, and normalized to β-actin (n = 4). (G) The protein level of FGF21 in culture medium of primary hepatocytes after stimulated with LPS for 12 h (n = 4). (H) Real-time PCR analysis for fgf21 in primary hepatocytes with LPS stimulation for 12 h (n = 3). (I) The expression of FGF21 in BMDMs after stimulated with LPS for various periods of time, and normalized to β-actin (n = 4). (J) The protein level of FGF21 in culture medium of BMDMs after stimulated with LPS for 12 h (n = 4). (K) Real-time PCR analysis for fgf21 in BMDMs with LPS stimulation for 12 h (n = 3). All values are presented as mean ± SEM. Statistical significance was measured using the unpaired 2-tailed Student t test for two experimental groups and one way ANOVA test for multiple groups. ns = no significant. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Next, we analyzed expression and secretion of FGF21 in hepatocytes, macrophages, and hepatic stellate cells in an inflammatory state to explore the origin of FGF21 in the liver of septic mice. To our surprise, in situ hybridization of septic liver tissue revealed that FGF21 was expressed not only in hepatocytes traditionally, but also in macrophages (Fig. 1(D, E)). Consistent with the in vivo results, besides upregulating FGF21 in hepatocytes (Fig. 1(F-H)), LPS promoted expression and secretion of FGF21 in bone marrow-derived macrophages (BMDMs) (Fig. 1(I-K)), but decreased FGF21 expression in LX-2 cells (a hepatic stellate cell line) (Fig. S1 (C-E)). These results suggest that hepatocytes and macrophages express FGF21 in the inflammatory state, which might modulate the progression of sepsis.

FGF21 protects against inflammatory septic liver injury induced by CLP

As previous study suggested that FGF21 KO mice have a lower survival rate than WT mice after LPS treatment [16], therefore, we investigated whether treatment with recombinant human FGF21 (rhFGF21) protects mice against CLP-induced mortality. As expected, in a short-term survival CLP model, rhFGF21 administration significantly reduced mortality from 100 % to 50 % in the first 24 h and improved the survival rate compared with saline-treated mice (Fig. 2A). To further elucidate the protective effect of FGF21 against sepsis, we examined whether rhFGF21 treatment protects mice against organ failure during sepsis. Analysis of lung morphology and H&E staining showed that rhFGF21 could not relief lung injury or prevent inflammatory cell infiltration in lungs of sepsis mice (Fig. S2(A-C)). This conclusion was further supported by the findings that the cell count and protein content of bronchoalveolar lavage fluid (BALF) did not differ between saline-treated mice and rhFGF21-treated mice after CLP (Fig. S2(D, E)). Consistent with it’ s lack of an effect on septic lung injury, rhFGF21 also could not alleviate renal tubular injury (tubular vacuolization and occasional detachment) or reduce the levels of blood urea nitrogen (BUN) and creatinine in serum of sepsis mice, indicating that rhFGF21 could not protect septic kidney injury (Fig. S2(F-H)). In contrast with the lungs and kidneys, analysis of terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining revealed that rhFGF21 alleviated CLP-induced cell apoptosis of liver (Fig. 2(B, C)). And H&E staining revealed that rhFGF21 reduced the number of infiltrated inflammatory cells in periphery of small and large vessels of liver in sepsis mice (Fig. 2(D-F)). Immunofluorescence staining of myeloperoxidase (MPO) showed that rhFGF21 restrained the recruitment of neutrophils in the liver of sepsis mice (Fig. 2(G, H)). Analysis of plasma biochemical parameters (alanine aminotransferase [ALT] and aspartate aminotransferase [AST]) also revealed that rhFGF21 mitigated CLP-induced liver dysfunction (Fig. 2I). Next, we determined the effect of FGF21 on the inflammation. As expected, the serum levels of inflammatory cytokines (IL-1β, IL-6, and TNF-α) and the mRNA levels of inflammatory cytokines in liver (cxcl1, cxcl2, il-1β, il-6, and tnf-α) were downregulated in sepsis mice treated with rhFGF21 (Fig. 2J and Fig. S2I).

Fig. 2.

FGF21 protects mice suppresses inflammation to against septic liver injury. (A) Short-term survival CLP model (n = 8). (B, C) Representative image and quantification of hepatic TUNEL staining of mice in indicated groups (n = 5). Scale bars: 50 μm. (D-F) Liver tissue samples from indicated groups were subjected to H&E staining, and infiltrated cells number around large vessel and small vessel was counted (n = 5). Scale bars: 50 μm. The region enclosed by the dashed lines represents the area infiltrated by inflammatory cells. (G, H) Representative image and quantification of hepatic MPO staining of mice in indicated groups (n = 5). Scale bars: 50 μm. (I) The Serum levels of ALT and AST from mice in indicated groups (n = 5). (J) Elisa analysis for IL-1β, IL-6, and TNF-α in serum of mice in indicated groups (n = 5). (K) The Serum level of FGF21 from mice in indicated groups (n = 5). (L) The expression of FGF21 in liver lysates from wildtype mice and FGF21 KO mice, and normalized to β-actin (n = 4). (M, N) Representative image and quantification of hepatic TUNEL staining of mice in indicated groups (n = 5). Scale bars: 50 μm. (O-Q) Liver tissue samples from indicated groups were subjected to H&E staining, and infiltrated cells number around large vessel and small vessel was counted (n = 5). Scale bars: 50 μm. The region enclosed by the dashed lines represents the area infiltrated by inflammatory cells. (R, S) Representative image and quantification of hepatic MPO staining of mice in indicated groups (n = 5). Scale bars: 50 μm. (T) The Serum levels of ALT and AST from mice in indicated groups (n = 5). (U) Elisa analysis for IL-1β, IL-6, and TNF-α in serum of mice in indicated groups (n = 5). Data are presented as mean ± SEM of each group. Statistical difference in mortality were analyzed by log-rank test. And one-way ANOVA test is using to assess the statistical differences in multiple groups. ns = no significant.

In addition, we used FGF21 knockout mice (FGF21 KO mice) to further verify the protective effect of FGF21 against septic liver injury (Fig. 2(K, L). The results showed that the cell apoptosis, infiltration of inflammatory cells around vessels, recruitment of neutrophils in the liver and serum level of ALT and AST were aggravated in FGF21 KO mice after CLP (Fig. 2(M−T)). Correspondingly, the serum levels of inflammatory cytokines (IL-1β, IL-6, and TNF-α) and the mRNA levels of inflammatory cytokines in liver (cxcl1, cxcl2, il-1β, il-6, and tnf-α) in liver were further increased in FGF21 KO mice (Fig. 2U and Fig. S2J). Moreover, measurement of constant bacterial loads in blood and peritoneal lavage fluid (PLF) of sepsis mice revealed that FGF21 did not affect the bacterial reproduction in septic mice (Fig. S2(K, L)).

Next, we used AAV2/8-CAG-Fgf21 to further validate the protective effect of FGF21 in sepsis mice. Based on the successful overexpression of FGF21 in liver of mice (Fig. S3 (A-D)), we found that overexpression of FGF21 in liver alleviated the liver dysfunction, inflammatory cells infiltration and cells apoptosis in liver of sepsis mice (Fig. S3 (E-L)). Taken together, these data indicate that FGF21 protects against septic liver injury by suppressing inflammation.

Autophagic flux is impaired in proinflammatory macrophages of septic mice

Autophagy was reported to participate in proinflammatory activation of macrophages [29]. And consistent with previous study that there was abundant autophagosomes accumulation in proinflammatory macrophages [30], [31], [32], we found that accompanied with inflammatory activation, the number of LC3 puncta and the expression of p62 was increased in macrophages in liver of CLP-operated mice (Fig. 3(A-F)). And in vitro, LPS also upregulated the protein level of LC3-II and p62 in macrophages (Fig. 3G).

Fig. 3.

LPS impaired autophagic flux in macrophages. (A, B) Representative immunofluorescence and quantification of iNOS-F4/80 double positive macrophages in liver from mice in indicated groups (n = 5). DAPI (blue), iNOS (red), and F4/80 (green). White arrows indicate iNOS-F4/80 double positive macrophages. Scale bars: 50 μm. (C, D) Representative immunofluorescence and quantification of LC3 puncta in F4/80-positive macrophages in liver from mice in indicated group. DAPI (blue), LC3β (red), and F4/80 (green) (n = 5). Scale bars: 50 μm. (E, F) Representative immunofluorescence and quantification of p62-F4/80 double positive macrophages in liver from mice in indicated groups. DAPI (blue), p62 (red), and F4/80 (green) (n = 5). White triangles indicate p62-F4/80 double positive macrophages. Scale bars: 50 μm. (G) The expression of LC3 and p62 in lysates of BMDMs in “control-group” and “LPS-group”, and normalized to β-actin (n = 4). (H) The expression of LC3 and p62 in lysates of BMDMs in indicated groups, and normalized to β-actin (n = 4). And Bafilomycin A1 was added for last 6 h. (I, J) Representative confocal images and quantification of Raw264.7 transduced with Lenti-GFP-mCherry-LC3B (n = 10). Scale bars: 20 μm. (K) The expression of LC3 and p62 in lysates of BMDMs in indicated groups, and normalized to β-actin (n = 4). 3-MA was given as pretreatment for 12 h before LPS administration. (L, M) Representative confocal images and quantification of Raw264.7 transduced with Lenti-GFP-LC3B in indicated groups (n = 8). Scale bars: 10 μm. (N) The expression of LC3 and p62 in lysates of BMDMs in indicated groups, and normalized to β-actin (n = 4). U0126, SP600125 and SB203580 were pretreated to BMDMs for 12 h. (O, P) Representative confocal images and quantification of Raw264.7 transduced with Lenti-GFP-LC3B in indicated groups (n = 8). Scale bars: 20 μm. (Q, R) Representative confocal images and quantification of pHLys Red in indicated groups (n = 5). Scale bars: 50 μm. (J) Real-time PCR analysis for lamp1, ctsa, ctsb, and ctsf from BMDMs in indicated group. All values are presented as mean ± SEM. Statistical significance was measured using the unpaired 2-tailed Student t test for two experimental groups and one way ANOVA test for multiple groups. ns = no significant. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Next, we used Bafilomycin A1 (Baf A1) and a GFP-mCherry-LC3B double-tagged lentivirus (yellow dots representing merged GFP and mCherry fluorescence indicate autophagosomes in a defect degradation, red dots representing mCherry fluorescence indicate autolysosomes) to analyzed dynamic autophagic flux in macrophages. Combined with accumulation of yellow dots in proinflammatory macrophage, and Baf A1 exacerbated the accumulation of LC3-II without affecting the protein level of p62, we speculated that LPS might promote formation of autophagosomes and blocke their degradation simultaneously (Fig. 3(H-J)).

To verified whether LPS promotes formation of autophagosomes in macrophages, we pretreated macrophages with 3-methyladenine (3-MA) (A typical inhibitor of autophagosome formation in early stage). As indicated by decreased the protein level of LC3-II and the number of GFP-LC3B puncta, we found that 3-MA blocked LPS-induced LC3 conversion and autophagosome formation in macrophages (Fig. 3(K-M)). In view of that MAPK signaling pathway (a typical downstream signaling pathway of Toll-like receptor 4) participates in LPS-induced autophagy [33]; therefore, we explored its involvement in LPS-treated macrophage. Treatment with U0126 (an inhibitor of ERK/MAPK) and SP600125 (an inhibitor of JNK /MAPK), but not SB203580 (an inhibitor of MAPK/p38), greatly downregulated the protein level of LC3-II and the formation of GFP-LC3B puncta but exert no effect on protein level of p62 (Fig. 3(N-P)), demonstrated that ERK/MAPK and JNK/MAPK activities are required for LPS-induced autophagosome formation in macrophages.

Furthermore, we found that LPS impaired lysosomal function in macrophages, which led to defective degradation of accumulative autophagosomes, as reflected by the decreased fluorescence intensity of pHLys Red (a probe of lysosomal acidic pH detection, which aggregate in normal lysosomes, and the fluorescence intensity increases as the pH decreases.) and mRNA level of lysosomal genes (lamp1, ctsa, ctsb, ctsf) in LPS-treated macrophages (Fig. 3(Q-S)). In summary, these results suggest that LPS promotes autophagosomes formation via ERK/MAPK and JNK/MAPK signaling pathway and impairs lysosomal function simultaneously, which finally lead to defective autophagic flux in macrophages.

rhFGF21 resumes autophagic flux in proinflammatory macrophage

Previous studies suggested that FGF21 regulates autophagy in various cell types to protect wound healing, liver injury and osteoarthritis [34], [19], [35]. And restoring blocked autophagic flux can effectively inhibit the pro-inflammatory activation of macrophages [31], [32]. Accordingly, we investigated whether FGF21 regulate autophagic flux of macrophages to suppress inflammation in sepsis. Accompanied by much more iNOS-positive macrophages, the number of LC3 puncta and expression of p62 in macrophages were higher in the liver of CLP-operated FGF21 KO mice than in the liver of CLP-operated WT mice (Fig. S4(A-F)). Meanwhile, LPS also induced higher protein levels of LC3-II and p62 in fgf21 knockdown macrophages in vitro (Fig. S4G). On the contrary, rhFGF21 treatment greatly reduced iNOS-positive macrophages, the number of LC3 puncta and expression of p62 in macrophages from the liver of sepsis mice (Fig. 4 (A-F)). In vitro, the heatmap of inflammatory genes showed that rhFGF21 significantly reduced mRNA level of inflammatory genes (tnfrsf9, il-1b, csf2rb2, il7r) in LPS-treated macrophages (Fig. 4G). Also, rhFGF21 reduced the protein level of iNOS and the secretion of inflammatory cytokines (IL-1β, IL-6, and TNF-α) of LPS-treated macrophages, which abolished by Baf A1, indicated repressive effect of rhFGF21 on proinflammatory macrophages was relied on fluent autophagic flux (Fig. 4H and Fig. S4H).

Fig. 4.

rhFGF21 impaired autophagic flux in macrophages. (A, B) Representative immunofluorescence and quantification of iNOS-F4/80 double positive macrophages in liver from mice in indicated groups (n = 5). DAPI (blue), iNOS (red), and F4/80 (green). White arrows indicate iNOS-F4/80 double positive macrophages. Scale bars: 50 μm. (C, D) Representative immunofluorescence and quantification of LC3 puncta in F4/80-positive macrophages in liver from mice in indicated group. DAPI (blue), LC3β (red), and F4/80 (green) (n = 5). Scale bars: 50 μm. (E, F) Representative immunofluorescence and quantification of p62-F4/80 double positive macrophages in liver from mice in indicated groups. DAPI (blue), p62 (red), and F4/80 (green) (n = 5). White triangles indicate p62-F4/80 double positive macrophages. Scale bars: 50 μm. (G) Heatmaps and quantitative results of mRNA level of inflammatory genes in BMDMs. (H) The expression of iNOS, LC3 and p62 in lysates of BMDMs in indicated groups, and normalized to β-actin (n = 4). And Bafilomycin A1 was added for last 6 h. (I, J) Representative confocal images and quantification of Raw264.7 transduced with Lenti-GFP-mCherry-LC3B (n = 10). Scale bars: 20 μm. (K, L) Representative electron micrographs and quantification from BMDMs in indicated group. Scale bars: 0.5 μm. Black arrows indicate autophagosome, and red arrows indicate autolysosome (n = 5). (M, N) Representative confocal images and quantification of pHLys Red in indicated groups (n = 5). Scale bars: 50 μm. (O) Real-time PCR analysis for lamp1, ctsa, ctsb, and ctsf from BMDMs in indicated group. Data are presented as mean ± SEM of each group. And one-way ANOVA test is using to assess the statistical differences in multiple groups. ns = no significant. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Next, we used Baf A1 and a GFP-mCherry-LC3B double-tagged lentivirus to analyzed the effect of rhFGF21 on dynamic autophagic flux in proinflammatory macrophages. The results showed that rhFGF21 restrained the expression of LC3-II and p62 protein level, and decreased yellow dots along with increased the number of red dots in LPS-treated macrophages, which indicated rhFGF21 resumed the blocked autophagic flux in LPS-treated macrophages. Moreover, cotreatment with Baf A1 also abolished the effect of rhFGF21 on dynamic autophagic flux (Fig. 4(H-J)). In addition, transmission electron microscopy (TEM) showed that rhFGF21 promoted the formation of autolysosomes to clear accumulated autophagosomes in LPS-treated BMDMs (Fig. 4(K, L)). Together, these results suggest that FGF21 restores autophagic flux impaired by LPS in macrophages.

To investigated the molecular mechanism of FGF21 in autophagy regulation, BGJ398 (a typical inhibitor of FGFRs) was introduced. The repressive effect of rhFGF21 on the protein level of LC3-II and p62 was largely abolished by BGJ398 treatment, indicating that the regulative role of FGF21 in autophagy depends on the activation of FGFRs (Fig. S5A). As reflected by decreased the phosphorylation levels of mTOR and S6 along with unchanged phosphorylation states of ERK and JNK, we found that rhFGF21 inhibited mTOR signaling pathway (the typical regulator of autophagic flux) without affecting ERK signaling pathway and JNK signaling pathway (Signaling pathways participate in autophagosome formation in LPS-treated macrophages) (Fig. S5(B, C)). Next, we investigated whether rhFGF21 regulates autophagic flux in macrophage by inhibiting the mTOR signaling pathway. As expected, pretreatment of BMDMs with MHY1485 (an agonist of the mTOR signaling pathway) for 12 h abolished the effect of rhFGF21 on autophagic flux and inflammatory activation of macrophages as indicated by the increased protein levels of iNOS, LC3-II, and p62, and the accumulation of yellow dots (Fig. 4(I, J) and Fig. S5D). In view of the fact that the activity of mTOR signaling pathway has been found to be inversely regulated by AMPK signaling pathway (A typical downstream signaling pathway of FGF21) [29], [36]. Thereafter, we found that the repressive effect of rhFGF21 on the phosphorylation of mTOR in LPS-treated macrophages was abolished by Compound C (a specific antagonist of AMPK signaling pathway) (Fig. S5E).

In addition, we found that the fluorescence intensity of pHLys Red was increased by rhFGF21 in LPS-treated macrophages, thus the improved lysosomal function by rhFGF21 in proinflammatory macrophages was confirmed (Fig. 4(I, J)). Meanwhile, we found that rhFGF21 suppressed phosphorylation of TFEB (a typical transcription factor of lysosomal genes, which phosphorylated upon activation of mTOR signaling pathway and is sequestered in the cytoplasm to be degraded [37]), restored the decreased protein level of TFEB (Fig. S5G), and attenuated the repressed transcription of genes related to lysosomal genes (lamp1, ctsa, ctsb, ctsf) induced by LPS (Fig. 4O). Taken together, these data suggest that rhFGF21 rescues the impaired autophagic flux to inhibit proinflammatory activation of macrophages by downregulating the mTOR signaling pathway.

The protective effect of rhFGF21 on septic liver injury is abrogated in Atg7^△mye mice

To further confirm our hypothesis in vivo, we crossed Atg7 flox/flox mice with LyzM-Cre mice to generate Atg7^△mye mice (Fig. S6A). Consistent with the in vitro results, rhFGF21 could not inhibit the proinflammatory activation of macrophage nor downregulate the serum levels of inflammatory cytokines (IL-1β, IL-6, and TNF-α) and the mRNA level of inflammatory cytokines (cxcl1, cxcl2, il-1β, il-6, and tnf-α) in liver (Fig. 5(A-C) and Fig. S6B). Moreover, rhFGF21 could not relieve the aggravated infiltration of inflammatory cells, apoptosis of cells in the liver, or liver dysfunction in CLP-operated Atg7^△mye mice (Fig. 5(D-K)). These findings indicate that rhFGF21 restores autophagic flux in macrophages and thereby inhibits their proinflammatory activation, which ameliorates inflammatory liver injury and improves liver function.

Fig. 5.

The protective effect of rhFGF21 in septic liver injury is invalid in Atg7^△mye mice. (A, B) Representative immunofluorescence and quantification of iNOS-F4/80 double positive macrophages in liver from mice in indicated groups (n = 5). DAPI (blue), iNOS (red), and F4/80 (green). White arrows indicate iNOS-F4/80 double positive macrophages. Scale bars: 50 μm. (C) Elisa analysis for IL-1β, IL-6, and TNF-α in serum of mice in indicated groups (n = 5). (D) The Serum levels of ALT and AST from mice in indicated groups (n = 5). (E-G) Liver tissue samples from indicated groups were subjected to H&E staining, and infiltrated cells number around large vessel and small vessel was counted (n = 5). Scale bars: 50 μm. The region enclosed by the dashed lines represents the area infiltrated by inflammatory cells. (H, I) Representative image and quantification of hepatic MPO staining of mice in indicated groups (n = 5). Scale bars: 50 μm. (J, K) Representative image and quantification of hepatic TUNEL staining of mice in indicated groups (n = 5). Scale bars: 50 μm. Data are presented as mean ± SEM of each group. And one-way ANOVA test is using to assess the statistical differences in multiple groups. ns = no significant. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

FGF21 restrains proinflammatory activation of macrophages by downregulating HIF-1α to protect against septic liver injury

Lines of evidence have established a conclusion that increased protein level of HIF-1α in inflammatory state of immune cell was attributed to upregulated transcription of hif-1α accompanied by suppressed degradation of HIF-1α [38], [39]. Meanwhile, HIF-1α was reported to directly regulate transcription of inos and il-1β, and promote transcription of glycolytic genes (hk2, pkm2, ldha), which contributes to metabolic reprogram of immune cells in order to meet the energy requirement in inflammatory conditions [30], [40], [41]. Also, recent pathological retinal neovascularization study revealed that a long-acting FGF21 molecule (PF-05231023) inhibited HIF-1α activity [42]. Given the previously established correlation between FGF21 and HIF-1α, we investigated whether FGF21 inhibits proinflammatory macrophage activation by downregulating HIF-1α.

As expected, rhFGF21 treatment reduced the number of HIF-1α-positive macrophages in the liver of CLP-operated mice (Fig. 6(A, B)). Meanwhile, rhFGF21 treatment decreased the protein level of HIF-1α in Kupffer cells from CLP-operated mice (Fig. S7A). Also, rhFGF21 treatment downregulated the mRNA levels of glycolytic genes (hk2, pkm2, and ldha) in liver of CLP-operated mice (Fig. S7B). In addition, more HIF-1α-positive macrophages and higher mRNA levels of glycolytic genes (hk2, pkm2, and ldha) were observed in the liver of FGF21 KO mice after CLP (Fig. S7(C-E)).

Fig. 6.

rhFGF21 inhibits proinflammatory activation of macrophage via downregulating HIF-1α. (A, B) Representative immunofluorescence and quantification of HIF-1α-F4/80 double positive macrophages in liver from mice in indicated groups (n = 5). DAPI (blue), HIF-1α (red), and F4/80 (green). White triangles indicate HIF-1α-F4/80 double positive macrophages. Scale bars: 50 μm. (C) The expression of HIF-1α in lysates of BMDMs in indicated groups, and normalized to β-actin (n = 4). (D, E) Representative immunofluorescence and quantification of iNOS-F4/80 double positive macrophages in liver from mice in indicated groups (n = 5). DAPI (blue), iNOS (red), and F4/80 (green). White arrows indicate iNOS-F4/80 double positive macrophages. Scale bars: 50 μm. (F) Elisa analysis for IL-1β, IL-6, and TNF-α in serum of mice in indicated groups (n = 5). (G) The Serum levels of ALT and AST from mice in indicated groups (n = 5). (H-J) Liver tissue samples from indicated groups were subjected to H&E staining, and infiltrated cells number around large vessel and small vessel was counted (n = 5). Scale bars: 50 μm. The region enclosed by the dashed lines represents the area infiltrated by inflammatory cells. (K, L) Representative image and quantification of hepatic MPO staining of mice in indicated groups (n = 5). Scale bars: 50 μm. Data are presented as mean ± SEM of each group. And one-way ANOVA test is using to assess the statistical differences in multiple groups. ns = no significant. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Consistent with the in vivo results, rhFGF21 treatment downregulated the protein level of HIF-1α (Fig. 6C), restrained nuclear translocation of HIF-1α, and suppressed transcription of glycolytic genes (hk2, pkm2, and ldha) in LPS-treated macrophages (Fig. S7(F, G)). Pretreatment with cobalt chloride (a typical agonist of HIF-1α) abolished the repressive effect of rhFGF21 on proinflammatory activation of macrophages, indicating that rhFGF21 restrains proinflammatory macrophage activation by downregulating HIF-1α (Fig. S7(H, I)).

To further confirm this hypothesis, we intravenously injected AAV2/8-CMV-DIO-HIF-1α-P2A-EGFP-tWPA to LyzM-Cre mice to generate LyzM-Cre^DIO-HIF-1α mice (Fig. S8(A, B)). Three weeks after the injection, both the number of GFP-positive macrophages and HIF-1α-positive macrophages were significantly increased in the liver of LyzM-Cre^DIO-HIF-1α mice (Fig. S8(C, D)). After CLP, rhFGF21 administration could not reverse inflammatory activation of macrophages in liver of LyzM-Cre^DIO-HIF-1α mice induce by CLP (Fig. 6(D, E)). Meanwhile, the repressive effects of rhFGF21 on the serum levels of inflammatory cytokines (IL-1β, IL-6, and TNF-α), the mRNA levels of inflammatory cytokines (cxcl1, cxcl2, il-1β, il-6, and tnf-α) and glycolytic genes (hk2, pkm2, and ldha) in liver were also abolished in LyzM-Cre^DIO-HIF-1α mice (Fig. 6F and Fig. S8(E, F)). And as a consequence of failure in anti-inflammation, rhFGF21 administration also did not ameliorate the aggravated inflammatory cells infiltration, increased cell apoptosis or deteriorated liver function in CLP-operated LyzM-Cre^DIO-HIF-1α mice (Fig. 6(G-L) and Fig. S8(G-I)). Collectively, these data suggest that FGF21 protects against septic liver injury by restraining inflammatory macrophage activation via downregulating HIF-1α in macrophages.

rhFGF21 promotes p62-dependent autophagic degradation of HIF-1α in proinflammatory macrophages

To explore the detailed molecular mechanism by which FGF21 represses HIF-1α, we first investigated whether FGF21 regulates homeostasis of HIF-1α. In the inflammatory state induced by LPS, activation of the MAPK and the NF-κB signaling pathway contribute to expression of HIF-1α [43]. Meanwhile, excessive ROS (reactive oxygen species) produced by damaged mitochondria and accumulation of succinate in proinflammatory macrophages suppress the activity of the PHDs (prolyl hydroxylases) which regulate the proteasomal degradation of HIF-1α [41], [44]. To our interest, rhFGF21 did not affect the mRNA level of hif-1α in LPS-treated BMDMs, indicating that rhFGF21 could not affect transcription of hif-1α (Fig. S9A). Cotreatment with cycloheximide (CHX) to block protein synthesis showed that rhFGF21 accelerated the degradation of HIF-1α (Fig. S9B). Indeed, immunoblotting revealed that even upon cotreatment with MG132 (a proteasome inhibitor), rhFGF21 still downregulated the protein level of HIF-1α in LPS-treated BMDMs, demonstrating that the effect of rhFGF21 on HIF-1α degradation is not attributable to the ubiquitin–proteasome system (Fig. S9C).

Given that HIF-1α accumulated in proinflammatory macrophages accompanied by abundant autophagosomes accumulation, we analyzed whether rhFGF21 downregulates HIF-1α by modulating autophagy. Immunofluorescence staining showed that rhFGF21 increased the colocalization of HIF-1α with Lamp1 (Fig. 7(A, B)). By contrast, cotreatment with Baf A1 largely blocked the effect of rhFGF21 on HIF-1α degradation (Fig. 7C). Consistent with previous study reported that the autophagy-lysosome pathway is involved in the degradation of HIF-1α, with p62 serving as the autophagic adaptor for HIF-1α [45], in this study, we found that the binding between HIF-1α and p62 in BMDMs was decreased by LPS treatment while rescued by rhFGF21 treatment (Fig. 7D). To determine whether rhFGF21 promotes autophagic degradation of HIF-1α dependent on p62, we used short interfering RNA to knockdown p62 in BMDMs. As expected, the effect of rhFGF21 on HIF-1α degradation was blocked in BMDMs depleted of p62 (Fig. 7E). Subsequently, we further investigated the molecular mechanism by which FGF21 affects the binding of HIF-1α and p62. Previous studies suggested that ULK1 mediated phosphorylation of p62 at Ser403 site increases the affinity of polyubiquitinated proteins with p62 [46], [47], and the activation of ULK1 is negatively regulated by mTOR through phosphorylation at its Ser757 site [48]. Given the repressive effect of FGF21 on mTOR signaling pathway in proinflammatory macrophages, we investigated whether FGF21 affects the phosphorylation of ULK1 and p62. Indeed, rhFGF21 inhibited the phosphorylation of ULK1 at the Ser757 site, thereby restoring the phosphorylation level of p62 at the Ser405 site in proinflammatory macrophages (Fig. 7F). Consist with results in vitro, rhFGF21 administration could not reverse the elevation of HIF-1α-positive macrophages or increased expression of HIF-1α and its downstream glycolytic genes (hk2, pkm2, and ldha) in the liver of CLP-operated Atg7^△mye mice (Fig. 7(G-J)). These data suggest that rhFGF21 promotes the degradation of accumulated HIF-1α in proinflammatory macrophages by modulating p62-dependent autophagy without affecting its transcription or proteasomal degradation.

Fig. 7.

rhFGF21 promotes autophagic degradation of HIF-1α. (A, B) Representative immunofluorescence and quantification of colocalization of HIF-1α and Lamp1 in BMDMs from indicated group (n = 5). DAPI (blue), HIF-1α (red), and Lamp1 (green). Scale bars: 20 μm. (C) The expression of HIF-1α in lysates of BMDMs in indicated groups, and normalized to β-actin (n = 5). Bafilomycin A1 was added for the last 6 h. (D) BMDMs were treated as indicated, and the cell lysates were subjected to immunoprecipitation with HIF-1α or p62 antibody, followed by immunoblotting with the HIF-1α or p62 antibody. MG132 (10 μM) and Bafilomycin A1 (100 nM) were added for the last 6 h. (E) The expression of HIF-1α in lysates of BMDMs in indicated groups, and normalized to β-actin (n = 5). BMDMs were transfected with Si-scramle (20 nM) and Si-sqstm1 (20 nM) for 48 h, then BMDMs were treated with single LPS or LPS + rhFGF21 for 12 h (n = 4). (F) The expression of p-ULK1 (Ser757) and p-p62 (Ser405) in lysates of BMDMs in indicated groups, and normalized to ULK1 and p62 (n = 4). (G, H) Representative immunofluorescence and quantification of HIF-1α-F4/80 double positive macrophages in liver from mice in indicated groups (n = 5). DAPI (blue), HIF-1α (red), and F4/80 (green). White triangles indicate HIF-1α-F4/80 double positive macrophages. Scale bars: 50 μm. (I) The expression of HIF-1α, LC3 and p62 in Kupffer cells isolated from liver of mice in indicated groups (n = 5). β-actin was used as the standard for verifying equivalent loading. (J) Real-time PCR analysis for hk2, pkm2, and ldha in liver from mice in indicated groups. Data are presented as mean ± SEM of each group. And one-way ANOVA test is using to assess the statistical differences in multiple groups. ns = no significant. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

PF-05231023 protects mice against septic liver injury

PF-05231023, an engineered FGF21 analog comprising two modified FGF21 molecules linked to a humanized IgG1 monoclonal antibody backbone, has a longer half-life (t½) than wild-type FGF21 [49]. Given that it has powerful pharmacological effects in primates with Type 2 diabetes mellitus (T2DM) and was well-tolerated by T2DM patients in a clinical test, we further assessed the therapeutic capacity of PF-05231023 in sepsis [50], [51]. Consistent with the effects of rhFGF21, PF-05231023 downregulated the protein levels of iNOS in LPS-treated BMDMs (Fig. 8A). In the short-term survival CLP model, PF-05231023 reduced the mortality of mice and improved their survival rate compared with that of saline-treated mice (Fig. 8B). Also, PF-05231023 relieved upregulation of inflammatory cytokines, liver dysfunction, infiltration of inflammatory cells and increased proinflammatory activation of macrophages in the liver of CLP-operated mice (Fig. 8(C-K)). These results indicate that PF-05231023 has the potential for clinical sepsis therapy.

Fig. 8.

PF-05231023 protects mice against septic liver injury. (A) The expression of iNOS in lysates of BMDMs in indicated groups, and normalized to β-actin (n = 4). (B) Short-term survival CLP model (n = 8). (C) The Serum levels of ALT and AST from mice subjected to mild-grade CLP in indicated groups (n = 5). (D) Elisa analysis for IL-1β, IL-6, and TNF-α in serum of mice in indicated groups (n = 5). (E-G) Liver tissue samples from mice in indicated groups were subjected to H&E staining, and infiltrated cells number around large vessel and small vessel was counted (n = 5). Scale bars: 50 μm. The region enclosed by the dashed lines represents the area infiltrated by inflammatory cells. (H, I) Representative immunofluorescence staining of MPO in liver section from mice in indicated groups (n = 5). Scale bars: 50 μm. (J, K) Representative immunofluorescence of iNOS-F4/80 double positive macrophages in liver from mice in indicated groups (n = 5). DAPI (blue), iNOS (red), and F4/80 (green). White arrows indicate iNOS-F4/80 double positive macrophages. All values are presented as mean ± SEM. Statistical significance was measured using one way ANOVA test for multiple groups. ns = no significant. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Discussion

Characterized by overwhelming host systemic inflammatory and immune responses to severe trauma and infection, which cause life-threatening organ dysfunction, there are nearly 30 million cases of sepsis per year and overall mortality is up to 25 % worldwide [2], [3]. Liver dysfunction in the inflammatory state induced by severe sepsis is associated with poor clinical outcomes. Although there is growing interest in targeting the liver to defend against sepsis, few strategies are currently available to ameliorate the damage caused by this life-threatening complication [5], [52], [53]. Using the CLP model of sepsis, we provide abundant evidence that FGF21 protects against septic liver injury by relieving hyperactivation of inflammation and consequent liver dysfunction. Although rhFGF21 significantly reduced mortality from 100 % to 50 % during the first 24 h in a short-term survival CLP model, the survival rate of rhFGF21-treated mice only reached 12.5 %, indicating that rhFGF21 cannot completely protect mice against sepsis but prolongs the therapeutic window.

To maintain homeostasis, autophagy was reported to modulate inflammation in immune cells involved innate immunity. In previous studies reported that formation of autophagosomes is increased, and their degradation is delayed in macrophages in inflammatory condition, indicating that autophagy is activated in response to inflammatory stress as a self-protective effect, while autophagic flux is blocked with development of inflammation [31], [32], [54]. The current findings also establish a tight relationship between autophagy and pyroptosis (inflammatory necrosis) in macrophages in which autophagy inhibits inflammasome activation by degrading important components (AIM2, NLRP1, NLRP3 and pro-caspase1) of the inflammasome [55], [56], [57], [58]. In addition to directly inhibiting pyroptosis, autophagy mediates the clearance of impaired mitochondria that generate excessive reactive oxygen species (ROS) in proinflammatory immune cells to enhance inflammatory signaling [59]. Several lines of evidence show that ROS promote NF-κB signaling pathway and suppress the activity of PHDs to stabilize HIF-1α in the inflammatory state [38], [41]. As negative feedback, HIF-1α may activates mitophagy by promoting transcription of bnip3 to eliminate damaged mitochondria and thus prevent excessive production of ROS [60], [61]. In this study, we demonstrated that HIF-1α accumulates upon impairment of autophagic flux in macrophages to mediate their proinflammatory activation. In proinflammatory macrophages, previous studies provided strong evidence that activated NF-κB and MAPK signaling pathway promote upregulate the expression of HIF-1α, while excessive production of ROS contributes to stabilization of HIF-1α [62], [63]. Interestingly, we demonstrated that rhFGF21 inhibits the mTOR signaling pathway to resume autophagic degradation of HIF-1α and thereby relieves proinflammatory activation of macrophages without affecting the mRNA level of hif-1α or proteasome degradation of HIF-1α. Using LyzM-Cre^DIO-HIF-1α mice and Atg7^△mye mice, we validated that the protective effect of rhFGF21 against septic liver injury relies on modulation of autophagy in macrophages to degrade HIF-1α for the first time.

Dysfunction of multiple organ (lungs, kidneys, liver and heart) is common during progression of sepsis [2], [64], [65]. Although FGF21 improved septic liver injury, it did not relieve septic injury of the lungs or kidneys. These differences may be attributable to differences in the circulatory structure and cell composition in liver, lungs and kidneys. Despite a recent study suggested that hepatic FGF21 preserves thermoregulation to increase the tolerance of mice during bacterial inflammation, the potential of FGF21 to affect immune regulation and suppress inflammation was almost ignored in this research [66].

Due to their low or lack of affinity for heparan sulfate, endocrine FGFs (FGF15/19 subfamily), including FGF15/19, FGF21, and FGF23, can enter the circulation and act as hormones [12]. This property means that FGF15/19 subfamily have powerful potential in clinical settings. In addition to FGF21 reported in this study, a recent study found that, FGF15/19 inhibits inflammation and modulates the Treg response in sepsis mice [67]. Besides the endocrine FGFs, some paracrine FGFs (FGF2, FGF5, and FGF9) also have therapeutic potential in sepsis [68], [69], [70]. Although these studies suggested that FGFs correlate with progression of sepsis, their effects and the underlying molecular mechanisms have not been studied in detail. Meanwhile, little is known about the role of some anti-inflammatory FGFs (FGF1, FGF10, FGF18, FGF20, and FGF23) in sepsis therapy [71], [72], [73], [74], [75]. Moreover, we have only demonstrated the potential function of FGF21 in macrophages during sepsis, and it remains to be explored whether FGF21 affects endothelial cells and other immune cells (neutrophils, dendritic cells, T cells and B cells) in this disease.

Conclusions

In this study, we demonstrated that expression of FGF21 is upregulated in hepatocytes and macrophages in the liver of mice with sepsis. Further exploration revealed for the first time that FGF21 inhibits proinflammatory activation of macrophages by restoring the autophagic degradation of HIF-1α, which effectively suppresses inflammation to protect septic liver injury. Importantly, FGF21 KO mice were more vulnerable to sepsis-induced hyperactivation of inflammation in the liver which was significantly relieved by rhFGF21.This suggests that administration of rhFGF21 or augmentation of endogenous FGF21 production may be a promising therapeutic strategy for sepsis in clinical settings.

Credit author statement

Junjie Zhu: Conceptualization, Methodology, Data curation, Formal analysis, Investigation, Projiect administration, Validation, Visulization, Writing-original draft. Zhouxiang Jin: Conceptualization. Jie Wang: Investigation, Projiect administration, Methodology. Zhaohang Wu: Investigation. Tianpeng Xu: Investigation. Gaozan Tong: Data curation, Validation. Enzhao Shen: Data curation, Validation. Junfu Fan: Formal analysis. Chunhui Jiang: Formal analysis. Jiaqi Wang: Formal analysis. Xiaokun Li: Conceptualization, Funding acquisition, Resources, Writing-review & editing. Weitao Cong: Conceptualization, Funding acquisition, Resources, Supervision, Writing-review & editing. Li Lin: Conceptualization, Funding acquisition, Resources, Supervision.

Compliance with ethics requirements

The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki was approved by the review committee of the Second Affiliated Hospital and Yuying Children's Hospital of Wenzhou Medical University. All animals received humane care according to the criteria outlined in the “Guide for the Care and Use of Laboratory Animals”. Animal protocols were approved by the Ethics Committee of Wenzhou Medical University.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary material

The following are the Supplementary data to this article: