Targeting Endothelial Erk1/2-Akt Axis as a Regeneration Strategy to Bypass Fibrosis during Chronic Liver Injury in Mice

Abstract

Liver sinusoidal endothelial cells (LSECs) have great capacity for liver regeneration, and this capacity can easily switch to profibrotic phenotype, which is still poorly understood. In this study, we elucidated a potential target in LSECs for regenerative treatment that can bypass fibrosis during chronic liver injury. Proregenerative LSECs can be transformed to profibrotic phenotype after 4 weeks of carbon tetrachloride administration or 10 days of bile duct ligation. This phenotypic alternation of LSECs was mediated by extracellular regulated protein kinases 1 and 2 (Erk1/2)-Akt axis switch in LSECs during chronic liver injury; Erk1/2 was normally associated with maintenance of the LSEC proregenerative phenotype, inhibiting hepatic stellate cell (HSC) activation and promoting tissue repair by enhancing nitric oxide (NO)/reactive oxygen species (ROS) ratio and increasing expression of hepatic growth factor (HGF) and Wingless-type MMTV integration site family member 2 (Wnt2). Alternatively, Akt induced LSEC profibrotic phenotype, which mainly stimulated HSC activation and concomitant senescence by reducing NO/ROS ratio and decreasing HGF/Wnt2 expression. LSEC-targeted adenovirus or drug particle to promote Erk1/2 activity can alleviate liver fibrosis, accelerate fibrosis resolution, and enhance liver regeneration. This study demonstrated that the Erk1/2-Akt axis acted as a switch to regulate the proregenerative and profibrotic phenotypes of LSECs, and targeted therapy promoted liver regeneration while bypassing fibrosis, providing clues for a more effective treatment of liver diseases.

Graphical Abstract

Proregenerative Liver sinusoidal endothelial cells (LSECs) can be transformed to profibrotic phenotype during chronic liver injury. In this study, Lao et al. showed that this phenotypic alternation was mediated by Erk1/2-Akt axis switch in LSECs and inhibited by LSEC-targeted treatments, providing clues for an effective therapy of chronic liver injury and fibrosis.

Introduction

The liver has a great capacity for regeneration. However, chronic or overwhelming liver injury may cause aberrant healing (fibrosis), which often overrides liver damage repair and may progress to cirrhosis and even hepatocellular carcinoma.1, 2, 3 Recently, increasing evidence has indicated that fibroblast-like cells, which are features of liver fibrosis, are mainly derived from activated hepatic stellate cells (HSCs); therefore, HSC activation is considered a pivotal event leading to fibrosis.1, 4 This process is tightly modulated by dynamic interactions between HSCs and other liver cell types, and understanding the cellular crosstalk involved in balancing effective regeneration and dysfunctional healing is expected to improve liver disease therapy.⁵

Liver sinusoidal endothelial cells (LSECs) are known to contribute to liver damage repair. After partial hepatectomy or acute liver injury, hepatic growth factor (HGF) and Wingless-type MMTV integration site family member 2 (Wnt2) are rapidly released by LSECs to enhance hepatocyte growth and repair the injured liver.6, 7 However, proregenerative LSECs are difficult to control, because chronic or overwhelming liver injury may result in LSEC dysfunction and elicit a fibrotic response from these cells, restricting the clinical application of LSECs as targets for promoting repair of liver damage.⁶ Therefore, it is important to understand how LSECs regulate liver regeneration and fibrosis. Previous studies have shown that vascular endothelial growth factor (VEGF)-induced LSECs mediate liver regeneration,⁷ prevent HSC activation, and promote the resolution of liver fibrosis.8, 9, 10 However, VEGF also stimulates fibrogenesis and the associated angiogenesis during chronic liver injury,11, 12, 13 making the effects of LSECs on HSC status and liver fibrosis unclear. Thus, more in-depth research on the mechanisms of LSEC interconversion between proregenerative and profibrotic phenotypes and the causal links between LSEC phenotypes and HSC activation is required. HSCs and LSECs are both derived from mesoderm, and their physical proximity and shared expression of angiocrine factors4, 14 make it easier for them to interact with each other. Although differentiated LSECs prevent HSC activation as described above, there is minimal evidence to prove that the fate of inactivated HSCs is regulated by LSECs.2, 15, 16, 17, 18 In addition, the mechanism of HSC fate regulation by LSECs is poorly understood.

We recently discovered that LSECs maintained their proregenerative phenotype during the early phase of chronic liver injury in both carbon tetrachloride- (CCl₄) and bile duct ligation (BDL)-induced mouse models. By 6 weeks of CCl₄ administration or 10 days of BDL, LSECs switched to a profibrotic phenotype and initiated the fibrosis process. During this process, the extracellular regulated protein kinases 1 and 2 (Erk1/2)-Akt axis in LSECs acted as a switch, altering the nitric oxide (NO)/reactive oxygen species (ROS) balance in the LSECs and changing the LSEC functional phenotype. The mammalian target of rapamycin (mTOR), a downstream target of Akt, reduced Erk1/2 activity and played a role as a switch regulator. We also found that different LSEC phenotypes mediated divergent HSC fates; proregenerative LSECs maintained HSC quiescence, whereas profibrotic LSECs stimulated HSC activation and commitment to senescence. Cytokines and chemokines secreted by profibrotic LSECs, such as interleukin-10 (IL-10) and regulated on activation, normal T cell expressed and secreted (RANTES), were involved in HSC activation and senescence. Use of an LSEC-targeted adenovirus or a drug particle that regulates the Erk1/2-Akt axis alleviated liver fibrosis and accelerated fibrosis resolution, providing a novel strategy for antifibrosis therapy.

Results

A Switch in LSEC Phenotypes Occurred during Chronic Liver Injury

To determine whether LSECs exhibited a proregenerative phenotype during chronic liver injury, we detected the functional phenotypes of LSECs from mice with CCl₄- or BDL-induced liver injury. After administration of CCl₄ or BDL, levels of LSEC-derived HGF and Wnt2, which elicit a proregenerative phenotype6, 7 and antifibrotic response,19, 20, 21 were continuously increased during early injury (CCl₄: weeks 0–4; BDL: days 0–10) but decreased after that time (Figure 1A; Figure S1A). In addition, LSEC differentiation, determined by fenestration on the cell surface, linearly declined, and completely dedifferentiated LSECs were clear after early injury (Figure S1B). Meanwhile, previously described molecular markers of dedifferentiation (CD31, LaminB1, Caveolin-1, and Rac1)22, 23 were upregulated (Figure S1C). The liver injury-associated proteins serum alanine transaminase (ALT) and aspartate transaminase (AST) were effectively maintained normally during early injury until HGF and Wnt2 were downregulated (Figure 1B; Figure S1D). Ki-67 expression, PCNA signals in the liver parenchyma, and expression of other hepatocyte proliferation-associated proteins in isolated hepatocytes followed a trend similar to that of HGF and Wnt2 expression and ALT/AST serum concentrations (Figures 1C–1E; Figure S1E), suggesting that proregenerative LSECs played a transient regenerative role during early injury. Additionally, the liver fibrotic response, determined by the expression of alpha smooth muscle actin (αSMA), desmin, and other fibroblast-associated proteins (vimentin, collagen I, transforming growth factor β [TGF-β], tissue inhibitor of metalloproteinases 1 [TIMP1]), began to be induced once the amount of proregenerative LSECs was reduced (CCl₄: after week 4; BDL: after day 10) (Figures 1C–1E; Figure S1E). Therefore, we propose that profibrotic LSECs replaced proregenerative LSECs and promoted liver fibrosis when the amount of proregenerative LSECs was reduced. These findings indicated that the proregenerative LSECs that play a role in liver regeneration after acute liver injury, as described previously,⁷ also perform this function in early chronic liver injury.

Figure 1.

The LSEC Proregenerative Phenotype Switched to a Profibrotic Phenotype during Chronic Liver Injury

(A) Expression of HGF and Wnt2 in LSECs was analyzed by western blot and gray-scale quantitation at different time points during chronic liver injury from three independent experiments. Data are expressed as the mean ± SEM; the Student’s t test was used, **p < 0.05. (B) Concentrations of ALT and AST in serum at different time points after CCl₄ administration in mice. Data are expressed as the mean ± SEM; n = 5 in each group; and the Mann-Whitney U test was used, **p < 0.05. Representative images demonstrating a Ki-67⁺ cell (red), an HSC (αSMA, green), and a hepatocyte (HNF4α, blue) (C), and a PCNA⁺ cell (red), an HSC (desmin, green), and nuclear staining (DAPI, blue) (D) markers in liver tissues of mice in the CCl₄ model are shown. Scale bar, 100 μm. The positive area and numbers of positive cells were calculated, and data are expressed as the mean ± SEM; n = 5 in each group; and the Mann-Whitney U test was used, **p < 0.05. (E) Expression of hepatocyte proliferation (left) and HSC activation (right) markers at different time points during chronic liver injury in at least three independent experiments were analyzed by western blot.

The Erk1/2-Akt Switch in LSECs Was Turned on during Phenotype Alteration

To demonstrate the mechanism underlying LSEC functional phenotype alteration, we first studied VEGF. VEGF is an essential regulator contributing to LSEC function to elicit regenerative responses and further inactivate HSCs6, 7, 8, 9 (which was also confirmed by our data in an LSEC-HSC coculture model, as shown in Figures S2A and S2B), and exerts bidirectional effects on fibrosis to stimulate both fibrogenesis and fibrosis resolution during liver fibrosis.24, 25, 26, 27 VEGF expression in the liver was increased with accumulating injury insults and was dramatically enhanced when the amount of proregenerative LSECs was reduced, whereas VEGF receptor 2 (VEGFR2) expression increased in proregenerative LSECs and decreased in profibrotic LSECs (Figure 2A). These findings support that a switch of the VEGFR2-VEGF axis determined different downstream pathways that had multiple effects during the liver injury process. To confirm this hypothesis, we examined the phosphorylation and expression of the VEGF downstream targets extracellular regulated protein kinases 1 and 2 (Erk1/2), p38, Src, and Akt. Erk1/2 phosphorylation at Thr202/Tyr204 was consistent with the trend in VEGFR2 expression, and Akt phosphorylation at Thr308 was consistent with the VEGF expression trend (Figures 2A–2D). Erk1/2 phosphorylation peaked in VEGF-induced proregenerative LSECs and was much higher in those cells than in profibrotic LSECs in vitro, although Akt phosphorylation was increased in both profibrotic and proregenerative LSECs (Figure S2C). In addition, we also noticed that LSEC autophagy, which protects LSEC homeostasis and attenuates liver fibrosis,²⁸ was also induced by VEGF in vitro, as determined by CYTO-ID staining (Figure S2A). However, Atg7 expression and the microtubule-associated protein 1 light chain 3 II (LC-3II)/LC-3I ratio were reduced, and p62 expression was induced when Akt was activated, suggesting that an autophagy defect occurred when Erk1/2 switched to Akt (Figures 2E–2H). Based on these results, we supposed that p-Erk1/2 in LSECs may be stimulated to maintain LSEC homeostasis and protect the liver from the fibrotic response during early injury, whereas p-Akt was activated to cause LSEC dysfunction and liver fibrosis. A higher p-Erk1/2/p-Akt ratio might be required for the LSEC proregenerative phenotype and HSC inactivation.

Figure 2.

The Erk1/2-Akt Switch in LSECs Occurred during Phenotype Alteration

(A) Expression levels of VEGF in liver tissues and of its downstream targets in LSECs in CCl₄-induced groups were analyzed by western blot. (B) Gray-scale analysis of p-Erk1/2 and p-Akt was done in three independent experiments. A Student’s t test was used, **p < 0.05 versus the CCl₄ 4-week (w) group. (C) Expression levels of VEGF targets and their phosphorylation in LSECs in BDL-induced groups were analyzed by western blot. (D) Gray-scale analysis of p-Erk1/2 and p-Akt was done in three independent experiments. A Student’s t test was used, **p < 0.05 versus the BDL 10-day group. Protein expression levels of autophagy markers in the CCl₄-induced (E) and BDL-induced (G) model were analyzed by western blot. Gray-scale analysis of LC-3 II/LC-3 I ratios in the CCl₄-induced (F) and BDL-induced (H) models were done in three independent experiments. A Student’s t test was used, **p < 0.05 versus the CCl₄ 4-week or BDL 6-day groups.

The Erk1/2-Akt Axis Was Required for Balance of the Proregenerative and Profibrotic Phenotypes in LSECs and for HSC Inactivation/Activation In Vitro

To confirm our hypothesis, we used the Erk1/2 agonist phorbol 12-myristate 13-acetate (PMA), the Erk1/2 inhibitor PD0325901, the Akt agonist Sc-79, and the Akt inhibitor LY294002 to regulate Erk1/2 and Akt activity in LSECs in an LSEC-HSC coculture model (Figure S3A). We found that PMA and LY294002 maintained LSEC differentiation (Figures S3B and S3C), and the number of cells undergoing autophagy (CYTO-ID Green positive) was increased in isolated LSECs treated with PMA and LY294002 (Figure S3D). The balance of NO and ROS was then examined and was associated with endothelial cell homeostasis,²⁹ induced endothelial NO synthase (eNOS) expression, and reduced NADPH oxidase 2 (Nox2) expression (Figure S3E). Additionally, the NO ratio was increased, whereas the ROS ratio was decreased when LSECs were treated with PMA and LY294002 for 3 days8, 9 (Figure S3F), and less NO was released extracellularly compared with the amount of released ROS (Figure S3F), suggesting that ROS had effects on HSC instead of NO. LSECs treated with PD0325901 and Sc-79 exhibited results contrary to the above results (Figures S3B–S3F); these findings implied that induced Erk1/2 maintained LSEC homeostasis but that induced Akt stimulated LSEC dysfunction. Levels of HGF and Wnt2 were increased in LSECs by PMA and LY294002, which induced the LSEC proregenerative phenotype, but were decreased by PD0325901 and Sc-79 (Figures 3A and 3B). Moreover, when cocultured with proregenerative LSECs, hepatocytes expressed more 5-ethynyl-2'-deoxyuridine (EdU) than their counterparts did (Figures 3C and 3D). HSCs were consistently inactivated when cocultured with proregenerative LSECs but were activated when cocultured with the profibrotic ones (Figures 3E and 3F). These findings elucidated that enhancing the Erk1/2 ratio in the Erk1/2-Akt axis of LSECs not only maintained LSEC homeostasis but also elicited proregenerative responses and furthermore promoted HC proliferation and HSC inactivation.

Figure 3.

The Erk1/2-Akt Axis Was Required for Balance of the LSEC Proregeneration and Profibrosis Phenotypes and HSC Inactivation/Activation In Vitro

(A) Representative immunofluorescence images of HGF and Wnt2 expression in LSECs are shown. Scale bar, 20 μm. (B) The HGF- and Wnt2-positive areas were calculated; data are expressed as the mean ± SEM from three independent experiments, and a Student’s t test was used, **p < 0.05 versus the 3-day group, *p < 0.05 versus 3-day + VEGF group. (C) Representative immunofluorescence images of proliferation in hepatocytes cocultured with LSECs analyzed by EdU incorporation assay are shown. Scale bar, 50 μm. (D) EdU-positive cell numbers were calculated by ImageJ software; data are expressed as the mean ± SEM from three independent experiments, and a Student’s t test was used, **p < 0.05. (E) HSC activation was assessed by αSMA immunofluorescence under conditions of LSEC Erk1/2 and Akt activation. Scale bar, 100 μm. (F) The αSMA-positive area was calculated; data are expressed as the mean ± SEM from three independent experiments, and a Student’s t test was used, **p < 0.05.

HSC Fates Varied with LSEC Phenotypes

To further investigate the mechanisms of HSC inactivation mediated by proregenerative and profibrotic LSECs, we studied multiple HSC fates including apoptosis, autophagy, senescence, and quiescence16, 17, 18, 30 by coculturing HSCs with LSECs of different phenotypes. First, it was demonstrated that HSCs were negative for TUNEL (Figure S4), excluding the possibility that proregenerative LSECs induced HSC inactivation by apoptosis. Additionally, the lipid content and expression of peroxisome proliferator-activated receptor γ (PPARγ) and glial fibrillary acidic protein (GFAP) were maintained in HSCs, whereas cyclin D and cyclin E staining were decreased (Figure 4A), indicating that proregenerative LSECs did not promote apoptosis of activated HSCs but induced HSC inactivation by blocking activation of HSC-associated cycle and autophagy. Amazingly, senescence-associated beta-galactosidase (SA-β-gal) staining was positive, and levels of senescence-associated proteins (p16, γH2AX, and 53BP1) were increased in HSCs when cocultured with profibrotic LSECs (Figure 4B). These findings are different from the evidence supporting the theory that senescent HSCs limit liver fibrosis.¹⁶ However, senescent HSCs also accumulated during liver fibrosis progression to have a senescence-associated secretory phenotype, which may contribute to continuous inflammation, adjacent HSC activation, advanced fibrosis, and even liver cancer.³¹ To our knowledge, this report is the first to uncover that HSC senescence is mediated by other cell types.

Figure 4.

HSC Fates Were Determined by LSEC Phenotypes

(A) Representative images of quiescent HSC markers are shown. Scale bars, 100 μm (oil red O); 50 μm (immunofluorescence). The positive area was calculated; data are expressed as the mean ± SEM from three independent experiments, and a Student’s t test was used, **p < 0.05 versus the HSC d0-d6 group. (B) Representative images of senescent HSC markers are shown. Scale bars, 100 μm (SA-β-gal); 50 μm (immunofluorescence). The positive area was calculated; data are expressed as the mean ± SEM from three independent experiments, and a Student’s t test was used, **p < 0.05 versus the HSC d0-d6 group.

Secreted Factors Mediated LSEC-HSC Interactions

LSECs regulate HSC phenotypes by paracrine activities,8, 9, 10 which are involved in immunogenicity and inflammation in liver diseases.³² Here, we isolated primary LSECs from mice at NC, the 4th and 6th week of CCl₄ administration, and the 2nd and 4th week after CCl₄ cessation and cultured them for 24 hr to collect supernatant and analyze LSEC-derived cytokines. Most factors were increased by administration of CCl₄ for 6 weeks (Figure 5A; Table S3), which enhanced immunogenicity and induced fibrosis.33, 34 Several factors, including macrophage inflammatory protein 2 (MIP-2), IL-17, and monocyte chemoattractant protein-1 (MCP-1), which are considered profibrotic,35, 36, 37 had lower concentrations at the 4th week than those at NC (Figure 5A), indicating that the fibrotic response was slowed down by the weakened immunogenicity of proregenerative LSECs.

Figure 5.

Analysis of Secreted Factors from LSECs during Liver Fibrosis Progression and Resolution

(A) Heatmap of the concentrations of cytokines and chemokines secreted by LSECs analyzed from three independent experiments; n = 5 in each group. (B) RANTES and IL-10 serum concentrations at different time points during chronic liver injury were assayed by ELISA. n = 8 in each group; a Mann-Whitney U test was used, **p < 0.05 versus the NC group. Representative images of αSMA immunofluorescence (C) and SA-β-gal staining (D) in HSCs were obtained from three independent experiments. Scale bar, 50 μm (αSMA) and 100 μm (SA-β-gal). αSMA-positive area data and numbers of SA-β-gal-positive cells are expressed as the mean ± SEM from three independent experiments in each group, and a Student’s t test was used, **p < 0.05.

Two Akt-induced factors, RANTES and IL-10, which are separately associated with HSC activation³³ and senescence,³⁸ were both increased in serum at the 6th week (Figure 5B). To confirm their effects, neutralizing antibodies and recombinant RANTES and IL-10 were separately added into HSC supernatant derived from LSECs alone or LSECs cultured with LY294002 for 3 days. Recombinant RANTES (50 ng/mL) accelerated Akt-dependent HSC activation, whereas a RANTES-neutralizing antibody (0.3 μg/mL) inhibited this effect (Figure 5C). RANTES had less of an effect on HSC senescence (Figure 5D). An IL-10-neutralizing antibody (15 ng/mL) reduced Akt-dependent HSC senescence, but 0.6 ng/mL recombinant IL-10 promoted this process. However, neither recombinant IL-10 nor an IL-10 antibody had a significant effect on HSC activation compared with untreated HSCs (Figures 5C and 5D), suggesting that IL-10 was less involved in HSC inactivation than in HSC senescence.

The Erk1/2-Akt Axis in LSECs Was Required for Liver Regeneration and Fibrosis Balance In Vivo

To avoid any nonspecific effects of chemical Erk1/2 and Akt agonists and inhibitors, we used arginylglycylaspartic acid-roundabout guidance receptor 4 (RGD-ROBO4) modified adenovirus to separately express constitutively active mitogen-activated protein kinase (MEK)1 (Ad-CA MEK1) and dominant-negative Akt (Ad-DN Akt), which are specific for endothelial cells,39, 40 to investigate the role of Erk1/2 and Akt activity in LSECs during chronic liver injury. Most of the Ad-RGD-ROBO4 vectors were located in the liver sinusoidal endothelium (Figure S5A). More than 85% of LSECs were transfected by these vectors (Figure S5B), and these vectors had no effects on Erk1/2 and Akt phosphorylation of liver cell types besides LSECs (Figure S5D) after 6 weeks of CCl₄ administration (Figure S5C), suggesting that Ad-RGD-ROBO4 had a high degree of LSEC specificity.

To confirm the effects of Erk1/2 and Akt on LSEC phenotypes, we set up four treatment groups in the CCl₄ model and three treatment groups in the BDL model (Figures 6A and 6C). CCl₄ administration for 4 weeks and BDL for 10 days led to the turning point of the Erk1/2-Akt switch in LSECs as described above. Subsequently, Ad-RGD-ROBO4-CA MEK and Ad-RGD-ROBO4-DN Akt were administered to induce Erk1/2 phosphorylation and reduce Akt phosphorylation in LSECs to maintain LSEC homeostasis and differentiation (Figures S5D–S5F). The sinusoidal vascular density and CD31 expression in liver tissues were also reduced in the Ad-RGD-ROBO4-CA MEK and Ad-RGD-ROBO4-DN Akt groups (Figure S5G). These results suggested that induced Erk1/2 or reduced Akt play a role in differentiated LSEC maintenance and angiogenesis reduction. The LSEC proregenerative phenotype was induced in the Ad-RGD-ROBO4-CA MEK and Ad-RGD-ROBO4-DN Akt groups (Figure S5D), promoting hepatocyte proliferation and a regenerative response against injury stimuli insult during early injury. Fibrotic markers were also reduced significantly in cells treated with Ad-RGD-ROBO4-CA MEK and Ad-RGD-ROBO4-DN Akt compared with cells treated with Ad-EGFP (Figures 6B and 6D; Figures S5H–S5J). We also determined whether the RANTES and IL-10 concentrations in serum were affected by Erk1/2 or Akt activity alteration and found that the RANTES level was reduced in Ad-CA MEK- and Ad-DN Akt-treated cells, whereas that of IL-10 had not significantly changed in any group (Figure S5K). To figure out each function of Erk1/2 and Akt, we performed Ad-RGD-ROBO4-DN MEK and Ad-RGD-ROBO4-CA Akt treatment to separately reduce Erk1/2 phosphorylation and induce Akt phosphorylation in LSECs (Figure S5L). After 4 weeks of CCl₄, Ad-DN MEK1 reduced Erk1/2 activity in LSECs and attenuated the liver regenerative response compared with that of the control treatment, whereas Ad-CA Akt induced Akt activity in LSECs and stimulated the liver fibrotic response compared with that of the control treatment after 6 weeks of CCl₄ (Figures S5M–S5O), demonstrating that Erk1/2 in LSECs mediates liver regeneration and induces LSEC Akt-stimulated liver fibrosis. These results demonstrated that enhancing the Erk1/2 ratio by targeting the Erk1/2-Akt axis in LSECs not only strengthened the liver regenerative response against CCl₄ insults and BDL injury, but also alleviated liver fibrosis, improving the progression of wound healing at the early phase of chronic liver injury.

Figure 6.

The Erk1/2-Akt Axis in LSECs Was Required for the Balance of Liver Regeneration and Fibrosis In Vivo

Timescale of chronic liver injury and cell therapy in the CCl₄- (A) and BDL-induced (C) models. Representative images demonstrating a Ki-67⁺ cell (red), an HSC (αSMA, green), and a hepatocyte (HNF4α, blue) in liver tissues of mice in the CCl₄- (B) and BDL-induced (D) models are shown. Scale bar, 100 μm. The positive area and numbers of positive cells were calculated. Data are expressed as the mean ± SEM; n = 8 in each group; and a Mann-Whitney U test was used, *p < 0.05, **p < 0.01.

LSEC-Targeted Honokiol Promoted Liver Regeneration while Bypassing Fibrosis through Erk1/2-Akt Axis Regulation

Next, we explored whether an LSEC-targeting treatment could induce liver regeneration while bypassing fibrosis by regulating the Erk1/2-Akt axis. Honokiol, an Akt inhibitor that also promotes Erk1/2 phosphorylation as described previously,32, 41 was loaded into RGD-labeled polyethylene glycol (PEG) nanoparticles to target endothelial cells for liver fibrosis treatment.⁴² These particles were intraperitoneally injected into CCl₄-treated mice two times per week for 6 weeks. As expected, Erk1/2 and Akt phosphorylation levels had less change in other cell types (Figure S6A) than that in LSECs (Figure S6B), suggesting that these particles had high LSEC specificity. Honokiol treatment maintained LSEC differentiation and the proregenerative phenotype in CCl₄-induced mice at the 6th week (Figures S6B–S6D). Hepatocyte proliferation was induced (Figures 7A and 7B; Figure S6E), and serum concentrations of ALT and AST were decreased (Figure 7D). Fibrosis was also significantly alleviated in the honokiol-treated group (Figures 7A and 7C; Figure S6E). Akt-induced RANTES and IL-10 were also inhibited by honokiol treatment (Figure 7F). In addition, proregenerative LSECs were still maintained by honokiol treatment at the 10th week of CCl₄ administration (Figure S6F). However, there was no difference in the hepatocyte proliferation response and the ALT/AST serum concentrations between the control and honokiol-treated groups (Figures 7A, 7B, and 7D; Figure S6F), although honokiol still induced an antifibrotic response (Figures 7A and 7C; Figure S6F). These results suggested that honokiol still alleviated the fibrosis process in liver cirrhosis but had no effect on liver regeneration.

Figure 7.

Honokiol Alleviated Liver Fibrosis and Promoted Fibrosis Resolution

(A) Representative images demonstrating a Ki-67⁺ cell (red), an HSC (desmin, green), and nuclear staining (DAPI, blue) markers in liver tissues of mice in the CCl₄ model are shown. Scale bar, 100 μm. Numbers of Ki-67⁺ cells (B) and the desmin-positive area (C) were calculated. Data are expressed as the mean ± SEM; n = 8 in each group; and a Mann-Whitney U test was used, **p < 0.05. Concentrations of ALT and AST in serum at different time points of CCl₄ administration during liver fibrosis progression (D) and liver fibrosis resolution (E) are shown. Data are expressed as the mean ± SEM; n = 8 in each group; Mann-Whitney U test, **p < 0.01, *p < 0.05. (F) Concentrations of RANTES and IL-10 in serum during liver fibrosis progression and liver fibrosis resolution were assayed by ELISA. Data are expressed as the mean ± SEM; n = 8 in each group; and a Mann-Whitney U test was used, *p < 0.05.

To evaluate the effect of honokiol on liver fibrosis resolution, we administered RGD-PEG-honokiol particles after CCl₄ cessation for 4 weeks followed by CCl₄ administration for 6 weeks. Erk1/2 phosphorylation and eNOS expression in LSECs were significantly increased, whereas Akt phosphorylation and Nox2 expression were rapidly decreased (Figure S6G). LSEC differentiation and the proregenerative phenotype recovered faster in the honokiol treatment group than in the control group (Figures S6H and S6I), as did the induced hepatocyte proliferation, the reduced ALT and AST concentrations, and the decreased degree of liver fibrosis (Figures 7A–7C and 7E; Figure S6J). The reduction in the RANTES and IL-10 serum concentrations was also accelerated by honokiol treatment (Figure 7F). These findings illustrated that honokiol promoted fibrosis resolution and accelerated liver regeneration by strengthening Erk1/2 activity in LSECs.

mTOR Played a Role in the Erk1/2-Akt Switch

As described above, induced VEGF activates both Erk1/2 and Akt simultaneously, and Erk1/2 phosphorylation is reduced when Akt is significantly activated. These observations led us to wonder whether there is an Akt-related mechanism that regulates Erk1/2 activity. mTOR, a downstream target of Akt that plays a pivotal role in controlling Erk1/2 activity,⁴³ was gradually induced during chronic liver injury (Figure 8A) and was considered to be the candidate pivotal point in the Erk1/2-Akt switch. Rapamycin was administered to inhibit mTOR activity in the Akt-inducing conditions established by Ad-RGD-ROBO4-CA Akt. We found that rapamycin reduced mTOR phosphorylation, led to a higher level of Erk1/2 phosphorylation than that in untreated cells established by Ad-RGD-ROBO4-CA Akt, and did not significantly change Akt activity (Figure 8B). Additionally, cell proliferation signals in the parenchyma adjacent to the injured lesion were induced, fibrotic septa were attenuated, and serum concentrations of ALT and AST were reduced in the rapamycin-treated group (Figures 8C and 8D). These results demonstrated that Akt-mTOR is one of the mechanisms that inhibits LSEC Erk1/2 activity and liver regeneration when Akt phosphorylation is significantly induced and Erk1/2 switches to Akt.

Figure 8.

mTOR Inhibited Erk1/2 and Promoted Erk1/2-Akt Switch

(A) LSEC expression of p-mTOR and mTOR at different time points during chronic liver injury were analyzed by western blot in at least three independent experiments. (B) Erk1/2, Akt, and mTOR expression and phosphorylation in different experimental groups after 6 weeks of CCl4 were analyzed by western blot in at least three independent experiments. (C) Representative images demonstrating a PCNA⁺ cell (red), an HSC (desmin, green), and nuclear staining (DAPI, blue) markers in liver tissues of different experimental groups of mice after 6 weeks of CCl₄ are shown. Scale bar, 100 μm. The positive area and numbers of positive cells were calculated. Data are expressed as the mean ± SEM; n = 5 in each group; and a Mann-Whitney U test was used, **p < 0.05. (D) Concentrations of ALT and AST in serum of different experimental groups after 6 weeks of CCl4 are shown. Data are expressed as the mean ± SEM; n = 5 in each group; and a Mann-Whitney U test was used, **p < 0.01, *p < 0.05.

Discussion

Using both CCl₄- and BDL-induced chronic liver injury models, we determined the turning point when LSECs switch from a proregenerative to a profibrotic phenotype. Proregenerative LSECs played a pivotal role in maintaining liver regenerative responses and HSC inactivation, whereas profibrotic LSECs promoted HSC activation and commitment to senescence. The switch of the Erk1/2-Akt axis in LSECs was a key event in the transformation of LSEC phenotypes. Erk1/2 activity was induced to preserve an LSEC proregenerative phenotype by increasing HGF and Wnt2 expression, and furthermore by promoting HSC inactivation. On the other hand, increased Akt activity contributed to inducing an LSEC profibrotic phenotype via induction of ROS and immunogenicity, leading to HSC activation, commitment to senescence, and advanced liver fibrosis.

In acute liver injury, a proregenerative LSEC phenotype helps with liver regeneration as described previously.⁶ In this study, we found that proregenerative LSECs were still maintained at the early stages of chronic liver injury. Notably, this phenotype was strengthened during LSEC differentiation, but it could be switched to a profibrotic phenotype once LSECs were completely dedifferentiated, which differs from previously described outcomes.8, 9 We supposed that at the “differentiated” status, the proregenerative phenotype of LSECs more likely protected the liver against injury, but once LSECs became completely “dedifferentiated,” the profibrotic phenotype dominated. Thus, the state of the LSEC proregenerative or profibrotic phenotype depended on the “differentiated or dedifferentiated” background rather than the degree of LSEC differentiation.

VEGF, in relation to the switch of LSEC phenotypes, exhibits a biphasic nature in the progression of liver fibrosis; that is, VEGF induces fibrosis-associated angiogenesis³⁷ and promotes fibrosis resolution.8, 9, 19, 24, 25, 27 We supposed that the dual effects depended on different downstream targets of VEGF rather than VEGF itself. Among the targets, Erk1/2 and Akt played a role in maintaining the LSEC proregenerative phenotype and stimulating the LSEC profibrotic phenotype, respectively. These two targets were both activated in response to VEGF during liver injury, which led us to wonder how Erk1/2 switches to Akt. In this study, we found that rapamycin inhibition of the Akt downstream target mTOR stimulated Erk1/2 activity in LSECs when LSEC Akt was highly activated, leading to liver regeneration maintenance and fibrosis inhibition. These results implied that mTOR probably formed a feedback loop to reduce Erk1/2 activity when Akt activity was induced to override and substitute for the Erk1/2 signal in LSECs.

In this study, the concept of NO and ROS balance was introduced for a more rational explanation of LSEC homeostasis. We found that enhancing Erk1/2 activity altered the NO and ROS balance toward NO in proregenerative LSECs, but toward ROS in profibrotic LSECs. NO was released extracellularly at a very low level, whereas ROS, which have been reported to promote liver fibrosis and HSC activation,⁴⁴ were released at a higher level. Therefore, blocking Akt-induced ROS released by LSECs was supposed to be an effective therapy for inactivating HSCs.

We further demonstrated that HSC fates were determined by LSEC phenotypes. Unexpectedly, pro-fibrotic LSECs promoted HSC senescence. To our knowledge, this report is the first to uncover HSC senescence mediated by LSECs. Although senescent HSCs limit cell growth and will be phagocytized by immune cells,¹⁶ the rate of HSC senescence induced by persistent liver injury may outpace cell clearance. Thus, senescent cells will be accumulated during liver fibrosis, which may contribute to continuous inflammation, advanced fibrosis, and even liver cancer.³¹ Therefore, pro-fibrotic LSECs not only play a permissive role in HSC activation but also advance liver fibrosis by inducing HSC senescence. Immune cells such as Kupffer cells and NK cells may provide a novel anti-fibrotic strategy by promoting senescent HSC clearance combined with the traditional treatment.16, 45

Findings in this study provide a potential strategy of anti-fibrosis therapy. In this study, ROBO4-loaded adenovirus was used to package LSEC targeting adenovirus. We noticed that 3-kb upstream promoter of ROBO4 contains information for endothelial cell type-specific expression in the intact endothelium;⁴⁶ this observation was confirmed by our published proteome data, which suggest that LSECs have highest ROBO4 abundance in murine liver,⁴⁷ and endothelial cell-specific adenovirus was also created based on this demonstration.³⁹ Furthermore, RGD, which is considered an endothelial cell-specific polypeptide,⁴⁸ was used to modify these adenoviruses and PEG particles carried with honokiol, ensuring the endothelial cell specificity of adenovirus and nanoparticle. Moreover, adenovirus had higher gene transfer efficiency than lentivirus and more space to assemble LSEC-specific elements than recombinant adeno-associated virus (rAAV) to ensure the high LSEC targeting, which may act as a more effective vector involved in anti-fibrotic treatment. Meanwhile, we also noticed that IL-10 concentration had not reduced in Ad-CA MEK1 and Ad-DN Akt administration, suggesting that immunogenicity of adenovirus might disturb the secretion of some LSEC-derived factors. Furthermore, these healing responses to proregenerative LSECs had a definite chronergy for regenerating the injured part of liver, and there may be some mechanisms offsetting the effects of proregenerative LSECs in advanced liver fibrosis and leading to cirrhosis initiation.

Our study uncovered that the switch from Erk1/2 to Akt was indispensable for transforming proregenerative LSECs to pro-fibrotic LSECs. LSEC-targeted treatments will provide some novel strategies for liver fibrosis therapy in the future.

Materials and Methods

Reagent Information

The TUNEL kit was purchased from Roche (Mannheim, Germany). A Cyto-ID autophagy detection kit was purchased from Enzo Life Science (Farmingdale, NY, USA). A SA-β-gal kit was purchased from Biovision (Milpitas, CA, USA). A NO detection kit was purchased from Beyotime (Shanghai, China). A ROS detection kit was purchased from SenBeiJia Bio (Nanjing, China). Western blot (WB) secondary antibodies, goat anti-rabbit and goat anti-mouse antibodies, were purchased from Cwbiotech (Beijing, China). Immunofluorescence secondary antibodies, goat anti-rabbit Alexa 488 and tetramethylrhodamine (TRITC) secondary antibodies, were purchased from Abcam, and DyLight 405 was purchased from Jackson Laboratory. Primary antibody suppliers are detailed in Tables S1 and S2.

Generation and Administration of Recombinant Adenoviruses

Replication-incompetent RGD-ROBO4 adenovirus was improved as described previously.⁴⁰ An endothelial-specific ROBO4 promoter sequence was amplified using PCR and put in place of the Trcp sequence of the pENTR-Trcp-3XFlag vector. Next, a BGH-PA sequence was added to the pENTR- ROBO4-3XFlag vector as a transcription ending signal. The coding regions of EGFP, CA MEK1, DN MEK1, CA Akt, and DN Akt were amplified and cloned into the pENTR-ROBO4-3XFlag-BGH-PA vector. The HI-loop of adenovirus type 5 (Ad5) was sheared by CRISPR/Cas9 using two small guide RNAs (sgRNAs), 5′-CTAAACGGTACACAGGAAACAGG-3′ and 5′-CCACAACTACATTAATGAAATAT-3′; then the amplified RGD-4C fragment CDCRGDCFC was inserted into the HI loop of the fiber knob of the original Ad5 fiber between amino acid residues 546 and 547. The inserted sequences were confirmed by partial sequencing analysis. Recombinant viral genomes were packaged into virus particles following transfection of HEK293 cells using a Gateway (Invitrogen, USA) cloning system. Ad5-RGD-ROBO4 virus was propagated in HEK293 cells, purified,⁴⁹ and diluted by PBS. The viral particle (vp) concentration was determined by real-time qPCR as described previously.⁵⁰

Animal Studies

Male C57BL/6J mice (body weight 20–25 g) were obtained from Vital River (Beijing, China). To establish the CCl₄-induced chronic liver injury model, CCl₄ was administered intraperitoneally at 1 mg/kg twice per week for 10 weeks. Fibrosis resolution was realized by CCl₄ cessation for 4 weeks after the 6th week, as previously described.6, 16 To establish the BDL-induced chronic liver injury model, we subjected 8- to 10-week-old male C57BL/6J mice to a midabdominal incision 3 cm long under general anesthesia. The common bile duct was ligated in two adjacent positions approximately 1 cm from the porta hepatis. The duct was then severed by incision between the two sites of ligation, as previously described.⁶

Recombinant adenoviruses were diluted in PBS and administered at a dose of 1 × 10¹¹ vp per mouse in 200 μL of PBS via tail vein injection. For the CCl₄ model, mice injected with olive oil and Ad-EGFP or CCl₄ and Ad-EGFP after 6 weeks of CCl₄ administration were regarded as normal controls or fibrotic controls, respectively. For the BDL model, mice with sham surgery or injected with Ad-EGFP at the 10th day after BDL surgery were regarded as normal controls or fibrotic controls, respectively.

RGD-PEG-loaded honokiol, kindly provided by Prof. Yan Wu (National Center for Nanoscience and Technology), was injected intraperitoneally at 0.2 mg/kg with CCl₄ administration. Mice intraperitoneally injected with olive oil or olive oil with physiological saline were regarded as normal controls. These experiments were conducted in accordance with the guidelines of the Institutional Animal Care and Use Committee of the National Center for Protein Sciences, Beijing.

Primary Cell Isolation and Coculture

A two-step liver perfusion in situ with collagenase IV digestion combined with further DNase I digestion in vitro was adopted; then differential centrifugation and OptiPrep density gradient centrifugation, magnetic-activated cell sorting to isolate LSECs and HSCs, and finally flow cytometry were used synchronously to evaluate cell purity. Specifically, for control mice, 100 μL of a heparin sodium solution was intraperitoneally injected to prevent blood coagulation, and 60 μL of a pentobarbital sodium solution was used for anesthesia; then HBSS perfusion with 6 mL/min for 6 min was performed and quickly cut off postcava. Next, the digestive liquid containing collagenase IV diluted with HBSS was perfused with 5 mL/min for 5 min. The resulting liver was gently torn to pieces by surgical tweezers in digestive liquid preheated to 37°C for approximately 6 min; then the cell digestion was terminated. The cell suspension was filtered through a 200 μm mesh cell filter to remove tissue residues; then centrifugation at 50 × g was done to precipitate hepatocytes. The rest of the supernatant was centrifuged at 500 × g for 8 min to acquire nonparenchymal cells. Then the cell pellet was suspended in a 24% OptiPrep density gradient working solution; then 17%, 11.2%, and 8.4% OptiPrep density gradient working solutions and a DMEM solution were gradually added onto the top of the cell suspension followed by centrifugation at 1,400 × g for 20 min. HSCs were collected at the layer between 8.4% and 11.2%, and LSECs and Kupffer cells were at the layer between 11.2% and 17% in the density gradient. Then magnetic-activated cell sorting with CD146 was done to enrich LSECs, and isolated cells were further stained with CD146 for flow cytometry analysis. Antibody usage followed the manufacturer’s instructions, and at the same time, a blank and an isotype control were run. For the mice with liver fibrosis, the whole process was similar to that described above, but some procedures required fine-tuning. Specifically, the HBSS perfusion was performed for 8 min, and the digestive perfusion for up to 8 min for full digestion of fibrotic liver. Then the 8.4% OptiPrep density gradient was changed to 8% to enrich for HSCs with high purity. Finally, the antibody usage varied according to the number of cells. Primary LSECs and HSCs with high purity and viability were obtained (Supplemental Information). HSCs and LSECs were cocultured using a transwell-based coculture system (Corning, New York, NY, USA). Primary HSCs were cultured in 24-well plates, and LSECs were cultured in 0.3-μm transwell chambers in DMEM (HyClone, South Logan, UT, USA) supplemented with 10% fetal bovine serum (FBS) (GIBCO, Grand Island, NY, USA) and penicillin and streptomycin (100 U/mL), as described previously.8, 9 A culture condition of 5% CO₂ and 37°C in a cell incubator was employed.

Immunochemistry Analysis

For immunocytochemistry, HSCs and LSECs were cultured on coverslips and fixed with 4% formaldehyde at room temperature for 30 min. The cells were permeated with 1% Triton X-100 and blocked with 3% BSA; then the HSCs and LSECs were incubated with primary antibodies. Next, all of the slices were incubated with anti-rabbit Alexa 488- or TRITC-conjugated secondary antibodies. Images of at least three sections for each slice were observed using a Zeiss LSM 880 confocal microscope (Zeiss, Germany). The experiments were performed as three independent experiments, and a representative image from each group is shown.

For immunohistochemistry, 30% sucrose dehydrated liver samples were sectioned, fixed in 4% paraformaldehyde, and subsequently incubated in 0.25% Triton X-100 for cell permeabilization. Sections were blocked in 5% BSA for 1 hr and incubated with primary antibodies overnight at 4°C. Secondary antibodies were incubated for 45 min at 37°C in the dark. Images of at least three sections for each slice were observed using a Zeiss LSM 880 microscope. Antibody conditions and references are detailed in Tables S1 and S2. The experiments were performed with at least five mice in each group, and a representative image from each group is shown.

Western Blot Analysis

For western blots, cells and liver tissues were lysed in radioimmunoprecipitation assay (RIPA) buffer with protease inhibitors. After standing for 30 min, proteins were fully extracted from lysate using a scroll oscillator or a tissue homogenizer. Then 12,000 × g centrifugation was performed to collect supernatant. This step was followed by protein quantification using a BCA protein assay kit or gel electrophoresis; equal amount of sample protein was loaded on 10% or 12% SDS-PAGE gels and transferred to a normal control (NC) membrane using an electroblotting apparatus. The blots were blocked in skimmed milk buffer and were next incubated with detection antibodies at 4°C overnight. The blots were then incubated with the corresponding secondary antibodies followed by membrane wash for 5 min three times. Using a 3,3'-diaminobenzidine (DAB) horseradish peroxidase color development kit, the bands were visualized and quantified by ImageJ software. Antibody conditions and references are detailed in Table S3. The experiments were performed as three independent experiments, and a representative image is shown.

Cell Apoptosis Analysis

TUNEL analysis was described previously.⁵¹ In brief, HSCs were fixed by 4% paraformaldehyde for 30 min and permeated by 0.1% Triton X-100 for 15 min. After being washed in PBS, the sections were incubated in a mixture (200 mM potassium cacodylate, 25 mM Tris-HCl [pH 6.6], 0.2 mM EDTA, and 0.25 mg/mL BSA) containing 0.2 U/μL TUNEL staining buffer and fluorescein isothiocyanate-labeled deoxyuridine triphosphate at 37°C for 1 hr. After being washed in PBS, the sections were observed under a Nikon LV100D fluorescence microscope. The experiments were performed with three wells of cultured HSCs in each group, and a representative image from each group is shown.

Cell Autophagy Analysis

The Cyto-ID Green autophagy staining followed procedures described previously.⁵² The 10× assay buffer was warmed to room temperature and then diluted to 1× with 9 mL of deionized H₂O and 1 mL of the buffer. The Cyto-ID Green autophagy dye solution was prepared by mixing 2 μL of the dye, 1 μL of Hoechst 33342, and 1 mL of 1× assay buffer. Adherent HSCs at 50%–70% confluence were covered with Cyto-ID Green autophagy dye solution for 30 min at 37°C. Then cells were fixed with 4% formaldehyde and analyzed by a Zeiss LSM 880 confocal microscope. The experiments were performed with three wells of cultured HSCs in each group, and a representative image from each group is shown.

Cell Senescence Analysis

Hepatic stellate cell senescence detection via detection of SA-β-gal activity was performed as described previously.⁵³ Adherent cells were fixed with 0.5% glutaraldehyde in PBS for 15 min, washed with PBS supplemented with 1 mM MgCl₂, and incubated overnight at 37°C in PBS containing 1 mM MgCl₂, 1 mg/mL X-Gal, and 5 mM each of potassium ferricyanide and potassium ferrocyanide. Sections were observed under a Nikon LV100D microscope. The experiments were performed with three wells of cultured HSCs in each group, and a representative image from each group is shown.

Oil Red O Staining

Oil red O staining was performed according to the manufacturer’s instructions as described. In brief, HSCs were fixed by 4% paraformaldehyde for 30 min. After being washed in PBS, cells were stained by an oil red O staining solution for 10 min and differentiated in 70% ethanol until sections appeared transparent. Then nuclei were stained by Mayer’s hematoxylin followed by washing in water for 3 min. After sealing the pieces, sections were observed under a Nikon LV100D microscope. The experiments were performed with three wells of cultured HSCs in each group, and a representative image from each group is shown.

Measurement of NO

The generation of NO was determined by measuring the stable NO metabolites, i.e., total nitrites, in culture medium with a nitrite detection kit as described previously.⁵⁴ In brief, 100 μL of LSEC lysate or culture medium was mixed with 100 μL of Griess reagent in a 96-well plate. The nitrite concentration was determined by spectrophotometry at 540-nm wavelength from a standard curve (0–100 μmol/L) derived from NaNO₂. All results were derived from three independently performed experiments.

Measurement of ROS

The ROS level was measured with a commercial ELISA kit according to the manufacturer’s instructions. In brief, 5 μL of LSEC lysate or culture medium was mixed with 45 μL of dilution buffer (1:10 dilution) in a 96-well plate. After being washed, samples were incubated in 50 μL of HRP-conjugated reagents at 37°C for 60 min. After another washing step, 50 μL of Chromogen Solution A and 50 μL of Chromogen Solution B were mixed in each well and incubated at 37°C for 15 min. Finally, stop solution was added, and the ROS concentration was determined by spectrophotometry measurement at 450-nm wavelength from a standard curve (0–100 U/L) derived from a standard solution. All results were derived from three independently performed experiments.

Cytokine and Chemokine Assay

For cytokine analysis, cell suspensions were cultured in DMEM with 10% heat-inactivated FBS, 2 mM L-glutamine, and 0.05 mM 2-ME at a concentration of 1 × 10⁶ cells/mL for 24 hr before supernatant harvest and analysis. Cytokine profiles were determined using the Mouse Cytokine/Chemokine Magnetic Bead Panel protocol from the Milliplex Map Kit (Millipore, Billerica, MA, USA). In brief, cytokine and chemokine assay plates were washed with washing buffer, sealed, and mixed on an orbital plate shaker at room temperature for 10 min. After the addition of the samples or controls, samples were incubated at 4°C overnight on an orbital shaker with fluorescently labeled capture antibody-coated beads. After overnight incubation with capture antibodies, the contents were removed via the washing instructions. Biotinylated detection antibodies were then added to each well and incubated with samples at room temperature for 1 hr while shaking. After incubation, the well contents were removed as previously described, and streptavidin-phycoerythrin was added to each well at room temperature for 30 min. Samples were then washed and resuspended in sheath fluid. Plates were run on the Luminex MagPix machine, and data were collected using the Luminex xPONENT software. Analysis of the cytokine and chemokine median fluorescent intensity (MFI) was performed using the Milliplex Analyst software. All results were derived from three independently performed experiments.

Statistical Analysis

Data are presented as the mean ± SEM, and n indicates the number of animals or experimental repeats that were performed, as indicated in the figure legends. Statistical analyses, power calculations, and graphical representations were done with appropriate software (GraphPad Prism [GraphPad Software, La Jolla, CA, USA] or ImageJ [NIH, Bethesda, MD, USA]). For statistical comparisons, the following tests were used as indicated in the figure legends: one-way ANOVA unpaired or paired, Mann-Whitney U test, and Student’s t test. The interaction of the effect of treatment group and time in different tested scenarios was assessed with an analysis of response profiles. The significance level was set at p < 0.05.

Author Contributions

Y. Lao and Y. Li designed and conducted the experiments, acquired and analyzed the data, and wrote the manuscript; P.Z., W.L., B.Q., and Y. Lv conducted the experiments, and acquired and analyzed the data; L.T., P.Z., S.S., Y.W., and H.W. provided reagents, adenovirus, and RGD-labeled honokiol coated with PEG nanoparticles; H.Z., C.T., and A.S. designed and supervised the experiments; and Y.J. and F.H. designed the study, supervised experiments, analyzed data, and wrote the manuscript.

Conflicts of Interest

The authors declare no competing interest.