Combination of G-CSF and a TLR4 inhibitor reduce in ﬂ ammation and promote regeneration in a mouse model of ACLF

Background & Aims: Acute-on-chronic liver failure (ACLF) is characterised by high short-term mortality, systemic in ﬂ ammation


Introduction
Acute-on-chronic liver failure (ACLF) is characterised by multiorgan failure in hospitalised patients with acute decompensation of cirrhosis. 1 The 90-day mortality rates vary from 30-100% depending on the age, number of organ failures and the severity of systemic inflammation. Systemic and hepatic inflammation, 2 nonapoptotic cell death [3][4][5] and a lack of hepatic regeneration 6 are major factors determining the development, course and prognosis of ACLF. Damage-associated molecular patterns (DAMPs) of both nuclear [core histones, high mobility group box-1 (HMGB1)] and cytoskeletal (cytokeratin 18, K18) origin are released by injured tissues where cell death is prevalent. 5,7,8 Furthermore, circulating pathogen-associated molecular patterns (PAMPs), of which lipopolysaccharide (LPS) is a key player 9 are also increased as a result of infections and/or bacterial translocation from the gut into the systemic circulation. Binding of DAMPs and PAMPs to their cognate pattern recognition receptors, predominantly Toll-like receptor 4 (TLR4), triggers NF-jB-dependent transcription and secretion of pro-inflammatory cytokines, such as IL1b, IL6 and TNFa leading to a cascade of inflammation and cell death. 8,10 Targeted inhibition of TLR4 with a small molecule inhibitor, TAK-242, has been shown to reduce the short-term mortality of animals with ACLF 4 but its administration is associated with a reduction in hepatocyte regeneration. 11 Keywords: liver failure; stem cell therapy; systemic inflammation; regeneration; senescence. Lack of hepatic regeneration is a feature of ACLF and, indeed, hepatocyte proliferation is a positive predictor of survival. 12,13 Hepatocellular senescence has been hypothesised to inhibit hepatic regeneration in cirrhosis 14 and hepatic inflammation is thought to be the main cause of inhibited cell cycle progression. 13,15 To modulate hepatic regeneration in ACLF, granulocytecolony stimulating factor (G-CSF), which mobilises bone marrow-derived stem and immune cells has been studied extensively in patients with ACLF. In several small, underpowered studies in patients with decompensated cirrhosis, alcoholic hepatitis and ACLF, G-CSF was shown to improve survival and was associated with mobilisation of CD34 + stem cells. [16][17][18][19] However, these results could not be reproduced in a large, placebocontrolled multicentre study in Germany. 20 Additionally, about twice the number of G-CSF-treated patients developed ACLF during follow-up compared with the placebo group. Moreover, 7 serious adverse reactions were observed in the G-CSF-treated group; in 5 this was related to aggravated organ failure. 20 Previous single-centre trials, using various dosing regimens ranging from 5 to 26 days with doses of 5 lg/kg once daily to 300 lg twice daily, reported similar beneficial results independent of the treatment scheme. The fact that the 2 multicentre trials covered a similar spectrum with short high-dose treatment (15 lg/kg once daily 21 ) and rather low-dose long-term therapy (5 lg/kg 20 ) speaks against the argument that the lack of efficacy was dose or timing related.
This study was designed to test the hypothesis that the lack of beneficial effect of G-CSF was possibly due to worsening of systemic inflammation through simultaneous release of bone marrow-derived inflammatory cells. We further hypothesised that this increased inflammation was through a TLR4-dependent mechanism, as previously shown in a rat model of partial hepatectomy. 23 The aims of this study were to determine whether the addition of TAK-242, a TLR4 antagonist, would prevent G-CSF-related aggravation of liver injury and enable hepatic regeneration. We also aimed to better define the underlying mechanisms of the interplay between G-CSF and TAK-242.

Study design
This study explored the hypothesis in 2 mouse models of ACLF. All animals were included in the analysis unless specified in the figure legends. Animals were randomly assigned to experimental interventions, but no blinding was applied. All experiments were performed and reported according to the ARRIVE guidelines. 24 Animal models All animal experiments were performed in accordance with UK Home Office Animals (Scientific Procedures) Act 1986 (updated 2012) and a project license (No.14378) provided by the UK Home Office. After intervention, animals were monitored closely and prematurely sacrificed if pre-comatose and/or if the disease condition rapidly worsened. Animals in each group were terminated by exsanguination under general anaesthesia with isoflurane (2% isoflurane in oxygen, Piramal Healthcare, USA).
Carbon tetrachloride-LPS model Male C57BL/6 mice (body weight: 30 g±4 g; age:8-10 weeks) were gavaged with carbon tetrachloride (CCl 4 0.5 mg/ml dissolved in olive oildose 0.5 ml/kg bw) twice weekly for a total of 6 weeks to induce liver fibrosis, as described previously. 25 To mimic ACLF, LPS (Klebsiella pneumonia, Sigma, UK), dissolved in saline to a final concentration of 6.25 lg/ll was injected i.p to a final dose of 4 mg/kg. Therapeutic interventions with G-CSF (250 lg/kg, s.c.) and/or TAK-242 (10 mg/kg, i.p.) were started 1 hour after LPS injection and repeated either 22 hours after LPS injection in the 24-hour model or daily in the 5-day model, respectively. Animals were sacrificed 24 hours or 5 days after LPS injections and 2 hours after the last therapeutic intervention. Time points were chosen based on our previous studies where we observed that there was significant liver injury 24 hours after LPS injection in CCl 4 pre-treated animals. 4 Liver regeneration and hepatocyte proliferation reaches its maximum earliest after 72 hours, with re-established liver mass after 7 days, 26 so that the second time point was chosen at 5 days after LPS injection.
Vehicle injections were performed with olive oil for CCl 4 , saline for LPS, saline for G-CSF and citric acid/NMP/Captisol for TAK-242.
Vehicle injections were performed with olive oil for CCl 4 , saline for GalN, saline for G-CSF and citric acid/NMP/Captisol for TAK-242.

Sampling and storage
Blood samples were taken from the right heart. Lithium heparin plasma was centrifuged 2,500 rpm for 10 min and stored at −80 C for later analysis. All tissues (liver, brain, kidneys) were snap frozen in liquid nitrogen and stored at -80 C for further analysis. In addition, for histological assessments, a sample of the organs was formalin-fixed (10% neutral buffered saline, Leica Biosystems, UK) for 24 hours before paraffin-embedding.

Statistical analysis
The sample size was calculated based on the results provided by Theocharis et al. 27 showing an alanine aminotransferase (ALT) reduction from 3,674 U/L ± 450 to 2,450 U/L ± 225 after treatment with recombinant G-CSF in mice with thioacetamineinduced acute liver failure. We hypothesised conservatively an effect size of 15% in our model with alpha error of 0.05, providing a power of 80% for samples of 7 animals per group and 90% for samples of 10 animals per group. The sample size was confirmed by using data from an ACLF rat model of bile duct ligation and LPS injection where LPS injection increased ALT levels to 129.4 ± 33 U/L whilst TAK-242 pre-treatment reduced it to 66.2 ± 9.4 U/ L. 4 By using an alpha error of 0.05, we calculated that 8 animals per group were needed for a power of 80% and 12 animals per group for a power of 90%. Finally, we aimed at having 10 animals per group unless stated otherwise.
All statistical analyses were performed using SPSS 24 software (SPSS Inc., Chicago; IL). Group comparisons for continuous variables were performed by using Man-Whitney U test and for categorical variables by using Chi-squared test. If more than 2 groups had to be analysed, one-sided ANOVA with post hoc Tukey analyses for multiple comparison were used. A p value < − 0.05 was considered statistically significant. Graphs were prepared in Prism (GraphPad, USA) and figures compiled in Adobe Photoshop (Adobe Systems, USA).
For further details regarding the materials and methods used, please refer to the CTAT table and supplementary information.

G-CSF increases the mortality of rodents with ACLF
To understand the pathophysiological basis of the lack of benefit and possible deleterious effect of G-CSF treatment that was observed in the German clinical trial (GRAFT study), 20 we performed preclinical studies in murine models of ACLF. In the first model, mice were gavaged for 6 weeks with CCl 4 and then injected with LPS to induce liver injury, as described previously 4,25 (Fig. 1A). Six-week CCl 4 administration induced bridging fibrosis (Fig. 1D). This model mimics the typical features consisting of multi-organ injury in response to administration of LPS and pre-existing chronic liver injury. 4,25 Twenty-four hours after LPS injection these animals developed significant liver and kidney injury with high ALT and creatinine levels and extended areas of hepatocyte cell death (TUNEL staining) (Fig. 2, Fig. S1). Hepatic expression of CCL5, ICAM-1 and NGAL, as markers of inflammation and organ injury, also increased after LPS injection compared to vehicle (p <0.001) (Fig. S2). 4 Liver injury, measured by ALT levels and liver cell death (TUNEL), was decreased 5 days after LPS injection (Fig. S1).
One-hour after LPS injection, mice were treated with G-CSF subcutaneously (250 lg/kg) once daily for 5 consecutive days.
Twenty-four hours after the first G-CSF injection the number of mobilised CD45 + cells in the blood increased from a median of 321,900 cells/ml blood (CCl 4 +LPS) to 510,504 cells/ml (CCl 4 +LPS+G-CSF) (Fig. 2B). G-CSF therapy led to a mortality rate of 66% after 48 hours (Fig. 1B) and a high degree of liver fibrosis  this model, G-CSF aggravated ACLF-related mortality, which was associated with worsening of hepatic macrophage infiltration.

TLR4 inhibition by TAK-242 prevents G-CSF-related mortality and inflammation
As our previous studies have shown increased hepatocyte TLR4 expression being involved in organ sensitisation to LPS in cirrhosis and a reduction in the severity of inflammation with inhibition of TLR4 in models of ACLF, 3 we hypothesised that the deleterious effect of G-CSF in ACLF may be modulated favourably by TLR4 inhibition. 28 We therefore used TAK-242, a small molecule TLR4-inhibitor, with or without G-CSF in the CCl 4 -LPS model over 24 hours and 5 days ( Fig. 2A). Administration of TAK-242 together with G-CSF reduced the mortality rate from 66%, in G-CSF-treated animals, to 0% (Fig. S10). Liver cell death was measured by TUNEL staining 24 hours after LPS injection. The median relative TUNEL-positive area in liver tissue increased from 0.7% (range 0. Inflammatory response was assessed by measuring liver cytokine expression at both the mRNA and protein level. LPS injection led to >4-fold increase in hepatic TNFa mRNA expression, whereas treatment with G-CSF induced an increase in IL-6 mRNA expression level (>2 fold), which was markedly reduced by TAK-242, with or without G-CSF (Fig. 2D). G-CSF increased the liver protein expression of inflammatory markers such as CXCL9, LIF and Petraxin3/TSG-14 as shown by the protein array data from pooled liver lysates. These markers were reduced after TAK-242 injection (Fig. S2 (Fig. 2F). High numbers of circulating neutrophils and Ly6c-high monocytes occurred in the blood after G-CSF therapy. Their numbers decreased with TAK-242 therapy and the combination of TAK-242 and G-CSF led to an increase of Ly6c-low monocytes (Fig. 2E). LPS injection was also associated with stellate cell activation (aSMA expression), which was significantly abrogated by TAK-242 (Fig. S4).
Monocytes and macrophages are major sources of cytokine release in liver disease. 29 We therefore tested to what extent LPS and G-CSF modulate the cytokine response in PMA (phorbol 12myristate-13-acetate)-activated (macrophage-like) and naïve THP1 cells (monocyte-like) in vitro. The experiments showed that G-CSF incubation of PMA-activated THP1 cells and naïve THP1 cells prior to LPS stimulation resulted in an aggravated cytokine response, especially upregulation of IL6 mRNA expression (THP1macrophages+LPS [10 ng/ml]: 6.2-fold upregulation vs. LPS (10 ng/ml)+G-CSF (100 ng/ml): 6.7-fold upregulation; THP1mo-nocytes+LPS [10 ng/ml]: 60.2-fold upregulation vs. LPS [10 ng/ ml]+G-CSF [100 ng/ml]: 71-fold upregulation [ Fig. S5A]). Results were confirmed in primary peripheral blood mononuclear cells isolated from healthy donors and incubated with pooled plasma from 9 patients with ACLF (Table S3). These cells exhibited a strong upregulation of IL-1b after incubation with G-CSF prior to LPS, compared to cells incubated with LPS alone. TAK-242 prevented excess cytokine production in peripheral blood mononuclear cells with or without G-CSF (Fig. S5B). These in vitro findings are in keeping with hepatic IL6 levels, which increased 2.8-fold in G-CSF-treated animals (Fig. 2D). These changes, either with LPS alone or in combination with G-CSF, were completely prevented in the presence of TAK-242, clearly showing that TAK-242 prevents the overwhelming inflammatory response driven by LPS and G-CSF. The data help to explain the reduction in markers of organ injury in the ACLF animals treated with TAK-242/G-CSF.
Hepatocellular senescence is a feature of ACLF, which is aggravated by G-CSF but prevented by TAK-242 Although modulation of inflammation using TAK-242 abrogated the severity of liver injury in the ACLF animals treated with G-CSF and/or LPS, this data does not uncover the mechanism underlying failure of regeneration in ACLF. Hepatocellular senescence, an essentially state of cell cycle arrest, is a pathophysiological feature of end-stage liver disease that inhibits regeneration and adequate tissue repair, resulting in poor clinical outcomes. 14 Mediators of cell cycle arrest, such as p16 or p21, are expressed in response to tissue injury and transforming growth factor beta-1 (TGFb1) is a key mediator of the senescenceassociated secretory phenotype (SASP), where a paracrine spread of injury-independent cellular senescence is observed throughout the liver tissue. 30,31 We therefore evaluated the effect of LPS and interventions with G-CSF and TAK-242 on markers of senescence.
LPS injection in CCl 4 -treated animals led to a prolonged (up to 5 days post administration) upregulation of both p16 mRNA (1.9-  14. Hepatic protein levels of p21 were raised starting from vehicle = 0.33 to CCl 4 +LPS 24 hours = 0.8 and CCl 4 +LPS 5 days = 1.14. G-CSF exaggerated TGFb1 (OD 1.1) and p21 expression (OD 0.97) 24 hours after LPS injection. TLR4 inhibition with TAK-242 effectively reduced protein expression of p21 and TGFb1 at both time points (all individuals per group). Western Blot was repeated twice and a representative blot is shown here. (C) Liver immunohistochemistry demonstrated that p21 expression occurred predominantly in hepatocytes and could be prevented by TAK-242, with or without G-CSF (n = 3 per group). 24 hours after LPS injection, 2.8% ± 2.2 hepatocytes expressed p21, which was enhanced to 8.7% ± 3 by adding G-CSF. This effect diminished 5 days after therapy. TAK-242 added to G-CSF significantly reduced the number of p21-expressing (senescent) hepatocytes to 0.2% ± 0.1, 24 hours after LPS injection (p <0.001). Image J was used for image quantification and group comparison was performed by one-way ANOVA and post hoc Tukey's multiple comparison only among the following groups: CCl 4 +LPS, CCl 4 +LPS+G-CSF, CCl 4 +LPS+TAK-242, CCl 4 +LPS+G-CSF+TAK-242. p >0.05 was considered statistically significant. mRNA data is delineated as DDCt value and a greater than 2-fold change in expression was considered as biologically relevant. Western blots were performed with protein lysates pooled from all animals per group and results shown in optical densitiy (OD). ACLF, acute-on-chronic liver failure; CCl 4 , carbon tetrachloride; G-CSF, granulocyte-colony stimulating factor; LPS, lipopolysaccharide; OD, optical density. (This figure appears in color on the web.) fold compared to vehicles) and TGFb1 in the liver (Western Blot, optical density [OD]: vehicle = 0.67 vs. CCl 4 +LPS 24 hours = 0.99 vs. CCl 4 +LPS 5 days = 1.14). Additionally, hepatic protein levels of p21 (Western Blot: OD vehicle = 0.33 vs. CCl 4 +LPS 24 hours = 0.8 vs. CCl 4 +LPS 5 days = 1.14) and Serpin E1/PAI-1 (proteome profiler) (Fig. 3A/B; Fig. S2C) were also upregulated by LPS, indicating the importance of senescence in this model of ACLF. Liver expression of both markers (p21 and Serpin E1/PAI-1) were further exacerbated by G-CSF treatment (p21 Western Blot OD CCl 4 +LPS+G-CSF 24 hours = 0.97). The administration of TAK-242, with or without G-CSF, effectively mitigated against the expression of all senescence markers (p16, TGFb1, p21 and Serpin E1/PAI-1) (Fig. 3A/B; Fig. S2C), as shown by mRNA expression (qPCR) and protein expression from pooled liver lysates. p21 immunohistochemistry revealed that activation of the p53/p21 pathway occurred predominantly in hepatocytes. The number of p21-expressing hepatocytes increased after 24 hours from 2.8% ± 2.2 with LPS alone to 8.7% ± 3 with LPS+G-CSF although this  Journal of Hepatology 2022 vol. 77 j 1325-1338 effect diminished 5 days after therapy. TAK-242 added to G-CSF significantly reduced the number of p21-expressing (senescent) hepatocytes to 0.2% ± 0.1, 24 hours after LPS injection (p <0.001) (Fig. 3C). Therefore, the data suggest that LPS is a key driver of hepatocellular senescence in ACLF, which is further exacerbated by G-CSF. TLR4 inhibition with TAK-242 prevents this senescent phenotype.
Failure of regeneration in ACLF is synergistically abrogated by G-CSF and TAK-242 G-CSF has been shown to exhibit regenerative properties in noninflammatory (not endotoxin-driven) models of liver injury. 32, 33 We therefore hypothesised that the excessive inflammation in ACLF, driven by LPS and/or G-CSF as observed above, may prevent G-CSF from acting as pro-regenerative agent and that the addition of TAK-242, which modulates inflammation, may overcome this.
To provide further evidence that G-CSF exhibits a proproliferative effect in an environment with lower grade inflammation, we tested 5 days of G-CSF therapy after CCl 4 administration without LPS. In this setting, G-CSF injection significantly increased the number of proliferating hepatocytes; Ki67 staining was observed in 4.1% ± 2.9 (p <0.001 to CCl 4 +LPS+G-CSF 5 days) and Cyclin A2 in 2.1% ± 1.3 (p <0.05 vs. CCl 4 +LPS+G-CSF 5 days) (Fig. S6).
Protein expression of other regenerative markers was assessed in pooled liver lysates using the proteome array. TAK-242 alone, administered over 24 hours in the CCl 4 -LPS mice, reduced both hepatic markers of vascular regeneration (such as angiopoietin 2, proliferin and platelet-derived growth factor) and other markers involved in liver regeneration, such as IL-22, Flt3-ligand and IGFBP-1 (Fig. S2D&E). Adding G-CSF to TAK-242 markedly increased the liver protein expression of these proregenerative markers (Fig. S2D,E).
These findings support the hypothesis that G-CSF requires an environment with reduced inflammation to exert its proliferative effects on hepatocytes and, therefore, TAK-242 allows G-CSF to exert its pro-regenerative capacities.

TAK-242/G-CSF reduces tissue injury in CCl 4 -GalNtreated mice
The effect of G-CSF, with or without TAK-242, was then tested in a second model where GalN, instead of LPS, was used as a sterile toxic insult in mice treated for 6 weeks with CCl 4 (Fig. 6).
Taken together these data suggest that in this model with low-grade inflammation, the combination of G-CSF and TAK-242 impacts positively on liver injury, regeneration and reduces markers of senescence. It also confirms that the regenerative effect of G-CSF is preserved in an environment with reduced inflammation.

Relationship between cell death and liver regeneration and the effect of G-CSF and TAK-242
To explore the role of cell death as a modulator and inducer of regeneration, we used RIPA56, a selective RIPK1 inhibitor, to prevent GalN-induced necroptotic cell death in the different groups of animals with CCl 4 -induced chronic liver injury. RIPA56 treatment effectively prevented GalN-driven cell death (total cell death [TUNEL p <0.001] and necroptosis [RIPK3 p <0.001], both compared to CCl 4 +GalN) and was associated with a decrease in hepatocyte proliferation compared to CCl 4 +GalN (Ki67 2% ± 1.7 vs. 0.2% ± 0.2, p <0.001; Cyclin A2 1.9% ± 1.6 vs. 0.03% ± 0.1, p <0.001) (Fig. S7), which suggests that in this environment with low-grade inflammation, the regenerative response correlates directly with the severity of liver injury, as previously observed. 30 We then calculated the ratio between expression of Cyclin A2, as a marker of cell cycle progression, and the severity of total liver cell death (TUNEL) to delineate the pro-regenerative capacity in relation to liver injury of both drugs in all models studied; an increase in the ratio of Cyclin A2/TUNEL would represent enhanced regenerative activity with less cell death and vice versa (Fig. S8). In the 24-hour CCl 4 -LPS model, LPS completely inhibited regenerative responses (CyclinA2/TUNEL ratio: CCl 4 0.42; CCl 4 +LPS 0.02; CCl 4 +LPS+G-CSF 0.04), with or without G-CSF. In contrast, in the CCl 4 -GalN model the regenerative response was preserved (CyclinA2/TUNEL ratio: CCl 4 +GalN 0.54). In both "short-term" models, administration of TAK-242 was associated with enhanced liver cell regenerative response (CyclinA2/TUNEL ratio: CCl 4 +LPS+TAK-242 1.4; CCl 4 +GalN+TAK-242 0.9) supporting the hypothesis that creating an 'inflammation-free' environment is important to restore regenerative capacities, as also described recently in a model of CCl 4 and Klebsiella pneumonia-induced ACLF. 15 Moreover, the anti-inflammatory environment (in the presence of TAK-242) allowed G-CSF to exert its positive effect on liver injury, notably in the CCl 4 -LPS model, and to enhance regeneration after 5 days of therapy (CyclinA2/TUNEL ratio: CCl 4 + LPS+TAK-242+G-CSF 1.7); thus, supporting the pro-regenerative capacity of G-CSF especially after long-term treatment.

Discussion
The results of this study provide novel insights into the mechanisms underlying the molecular pathogenesis of ACLF, focusing on the relationship between inflammation and regeneration. The data provide the rationale for combining G-CSF, a recombinant protein that mobilises stem cells from the bone marrow, and a TLR4 inhibitor, TAK-242, that has been shown to reduce hepatic inflammation and mortality in animal models of ACLF. Herein, we show for the first time that in models of LPS-or GalNinduced ACLF, G-CSF and TAK-242 act in combination to improve the severity of liver injury by reducing inflammation and cellular senescence and improving regeneration.
ACLF has a dismal prognosis, 34 and its treatment is an urgent unmet need. In ACLF, TLR4-driven inflammation and lack of hepatic regeneration determines the disease course. 4,12,15,35 Additionally, DAMPs, 36 released from non-apoptotic, immunogenic forms of cell death (such as necroptosis), and PAMPs, 3,4 derived from infection and/or intestinal bacterial translocation, both initiate a cascade of cytokine-driven inflammatory responses that underlie the pathogenesis of the syndrome in which the TLR4 pathway plays an important role. 4,5,8 This systemic inflammation is also associated with the release of endogenous G-CSF, which can induce further tissue injury. 2,35 Paradoxically, G-CSF has been used to treat patients with ACLF, with a view to mobilising the bone marrow stem cell with variable results. 18,19,21 Although preclinical studies suggested that G-CSF may help in tissue repair, 27 and some clinical trials in patients with ACLF showed promising results, 18 the large multicentre randomised clinical trial of G-CSF failed to show a beneficial effect. In fact, in the subgroup of patients with the most severe inflammation, there were some suggestions that G-CSF may even be deleterious. 20 In a rat model of septic liver injury, 4-day pre-treatment with G-CSF resulted in increased liver expression of LPS-binding protein and TLR4. Administration of LPS to these animals led to enhanced liver cell death and mortality showing a negative sensitising effect of G-CSF. 28 However, in human trials, administration of G-CSF to treat sepsis showed no evidence of efficacy but G-CSF did not induce any excess mortality. It is important to note that these patients had no underlying liver disease. 37 When rats were treated with G-CSF in a rodent model of partial (70%) = CSF, was superior to the single compounds. (C) The treatment effect was associated with activation of the STAT3 pathway and a trend to increased anti-apoptotic BCL2 expression as shown by Western blot in pooled samples. Western blot was repeated twice and a representative blot is shown here. (D) GalN induced a regenerative response with proliferating hepatocytes but also hepatocytes in cell cycle arrest. G-CSF alone or in combination with TAK-242 reduced cell death and the subsequent regenerative response. TAK-242 alone maintained the amount of proliferating (Cyclin A2, Ki67) and senescent (p21) hepatocytes. Image J was used for image quantification and group comparison was performed by one-way ANOVA and post hoc Tukey's multiple comparison only among the following groups: CCl 4 +LPS, CCl 4 +LPS+G-CSF, CCl 4 +LPS+TAK-242, CCl 4 +LPS+G-CSF+TAK-242. mRNA data is delineated as DDCt value and a greater than 2-fold change in expression was regarded as biologically significant. Western blots were performed with protein lysates pooled from all animals per group and results shown in optical density (OD). ACLF, acute-on-chronic liver failure; CCl 4 , carbon tetrachloride; GalN, galactosamine; G-CSF, granulocyte-colony stimulating factor; LPS, lipopolysaccharide. (This figure appears in color on the web.) hepatectomy, administration of LPS was associated with severe liver injury and excess mortality compared with partial hepatectomy and LPS injection alone. 23 These results indicate that G-CSF may sensitise the liver to LPS through TLR4, which may be deleterious in patients with ACLF, as these patients present with high levels of circulating DAMPs and PAMPs, potent ligands for this receptor. 5,9 To better understand the mechanisms underlying the potential deleterious effect of G-CSF, we used murine models of ACLF that mimic the human phenotype. Administration of LPS to mice with advanced fibrosis led to ACLF, 4,25 which was associated with severe inflammation, liver injury, hepatocellular senescence, and high mortality rate. G-CSF exaggerated the inflammatory response and further induced cell death by decreasing the BCL2/BAX ratio and hence favouring cell death. Inhibiting TLR4 receptor signalling with TAK-242, which was associated with marked reduction in the severity of systemic and hepatic inflammation, prevented the increased mortality with G-CSF. Furthermore, TAK-242 allowed G-CSF to act through activation of the STAT3 pathway to prevent further cell death by releasing BCL2 and inducing hepatocyte proliferation. These observations may indicate that the combination may indeed be synergistic in their action rather than simply additive. It may be argued that true recovery needs to be proven at multiple levels, which may be true especially for hepatotoxic injuries in acute liver injury. However, ACLF comprises a complex pathogenesis including inflammation, cell death and inadequate regeneration, with predominance of one or the other in different disease stages. We observed that the combinatorial therapy addresses all aspects and preserves tissue integrity exhibited by low ALT levels and 100% survival. G-CSF may also target side effects of TAK-242, which may impair response to pathogens such as phagocytosis increasing the risk of secondary infections. 38 However, TAK-242 did not increase the risk of infections in human trials and was shown to be safe in sepsis. 39 Nevertheless, any future trial will need to monitor closely for infections when patients are treated with a TLR4 antagonist.
ACLF is characterised by a lack of regeneration and LPS-driven inflammation is potentially of pathogenic importance. 15 In the LPS-precipitated model of ACLF, the number of Ki67 + hepatocytes were markedly reduced despite activation of Cyclin A2. This may be related to the inhibitory effect of hepatocellular senescence on hepatic regeneration, indicated by increased hepatocellular p21 expression. In human liver tissues of patients with ACLF, the presence of infiltrating CD68 + macrophages was associated with low numbers of proliferating hepatocytes 6 and in a ACLF mouse model, infection was an important determinant of inhibition of regeneration. 15 Taken together, the data support the hypothesis that targeting inflammation may improve regeneration.
The relationship between TLR4 signalling and cell proliferation is contradictory; previous in vitro work on tumour cell lines indicated a pro-proliferative effect on malignant parenchymal cells, [40][41][42] whereas TLR4 stimulation by LPS was proposed by others to halt pluripotent progenitor cells' ability to divide. 43,44 In cirrhotic animals, administration of LPS inhibited regeneration, as evidenced by activation of the IFNy-STAT1 pathway with increased expression of cyclin-dependent kinase complex inhibitor while reducing the expression of Cyclin D1. Additionally, administration of G-CSF to LPS-induced ACLF animals led to markedly increased expression of p21 in the hepatocytes, which was associated with enhanced expression of other senescence mediators such as Serpin1E, p16 and TGFb1. TLR4 signalling also leads to TGFb1 release, which is a classical mediator of the senescence-associated secretory phenotype, activating p21 independently of p53. 30,31 Taken together, senescence is an important consequence of LPS-induced ACLF, which is aggravated by G-CSF. These data allow us to hypothesise that senescence and the concomitant inflammatory environment may be involved in the inhibition of hepatic regeneration by senescence in ACLF. The fact that inhibiting the TLR4 pathway by TAK-242 in G-CSF treated animals reduces inflammation, prevents senscence and fosters regeneration further supports this hypothesis. Evidence for the co-existence of pathogen/endotoxin-driven inflammation and lack of regeneration is robust and alcoholic hepatitis, one of the leading precipitating events for ACLF, seems to be one of the potential indications for treatment with this combinatorial therapy. Dubuquoy et al. indicated that livers from patients with alcoholic hepatitis lack sufficient hepatocyte proliferation whilst showing increased proliferation of hepatic progenitor cells and ductular reaction. 12 The same disease is also characterised by high levels of circulating LPS and inflammation. 9 To further validate our observations and determine whether this pro-regenerative effect could be reproduced in an environment of non-LPS-driven, sterile injury, we created a new model of ACLF using GalN on the background of advanced fibrosis. We confirmed similar effects of the combination therapy in this model. Although G-CSF alone reduced tissue injury, its effect on tissue repair, when combining it with TAK-242, was superior to the individual treatments. This observation shows that TAK-242 overcomes the effect of injury-related release of DAMPs and TLR4 activation, confirming the beneficial effect of the combination therapy even when the superimposed injury is non-LPS related.
This study is limited by the fact that ACLF is a heterogeneous entity and animal models might mimic only part of the disease spectrum. However, using an inflammatory and a sterile model of ACLF, mimicking extrahepatic and hepatic insults, allowed us to explore the major pathogenic factors involved in disease development and progression. Although our study provides important insight into the inflammation-senescence-regeneration sequence and provides a potential novel therapeutic strategy, further research is needed to understand the involvement of endothelial cells and other subsets of immune cells (such as macrophages), which might modulate regenerative responses; their involvement is likely as indicated in this study by changes in the markers of vascular regeneration, such as angiopoietin, proliferin and platelet-derived growth factor, possibly also acting through modulation of TGFb1 expression. 45,46 It may also be necessary to apply the combinatorial therapy in models where infection induces ACLF 15 and models of alcoholic hepatitis to test the TAK-242/ G-CSF combination in models mimicking the most frequent precipitating events. The models applied here target mechanistic aspects of pathogenetic changes, which occur throughout the whole spectrum of ACLF, providing the basis for translation into humans and for future mechanistic studies.
Although the clinical course of ACLF is not confounded by sex, mechanistic preclinical studies should ideally include both male and female animals. Herein, we chose to use male animals only because the model applied in this study was previously developed in male animals.
In conclusion, the results of this study provide new evidence for the importance of the LPS-TLR4 pathway in modulating the systemic and hepatic inflammation that is associated with ACLF, as well in inducing hepatocyte senescence and inhibiting 1336 Journal of Hepatology 2022 vol. 77 j 1325-1338 regeneration. We report the novel observation that the combinatorial therapy of TLR4 inhibition using TAK-242 and G-CSF reduces liver injury and improves hepatocyte proliferation through reduction in inflammation and senescence. This overcomes the inhibition of hepatic regeneration, a characteristic feature of ACLF. As both drugs are known to be safe, they can be repurposed and evaluated in combination for patients with ACLF.

Financial support
The German Research Foundation (DFG) (EN 1100/2-1) and Berlin Institute of Health (BIH) funded Cornelius Engelmann. This study was also received funding by a grant from the European Union Horizon 2020 research and innovation programme under grant agreement number 954096.