The effect of BMSCs in clinical treatment is highly dependent on their quality. However, the lack of standardized culture procedures and unique markers limits the consistency of BMSC characteristics. The initial step for BMSC standardization is the isolation process (38). Human bone marrow is mainly obtained from the iliac crest via aspiration in the presence of some anticoagulants, such as heparin sodium (23). There are several methods for isolating BMSCs from bone marrow. Traditional differential adhesion is based on the typical capacity of MSCs, such as adherence to plastic and sensitivity to enzyme digestion, and the culture medium is changed every 3-4 days to gradually achieve purification. Although this method is convenient and economical, it does not yield a homogeneous population of cells that contain other bone marrow subpopulations, such as endothelial cells, pericytes, leukocytes, and haematopoietic stem cells (39, 40). Another technique, called density gradient centrifugation, has also been proposed to precipitate different cells in the bone marrow according to size and density using gradient centrifugation solutions with a density of approximately 1.077 g/mL, low viscosity and low permeability, such as Ficoll, Ficoll-Paque, Percoll, and Lymphogre. After centrifugation and stratification, the greyish-white cloudy layer above the separation fluid was purified through differential adhesion. However, the lack of specific subsets and contamination of cell populations have limited its application (38, 41, 42). To improve the homogeneity of BMSC populations, advanced isolation methods, such as fluorescence-activated cell sorting (FACS) and magnetic-activated cell sorting (MACS), have been used for high-throughput screening of BMSCs via specific surface markers. FACS and MACS employ electric and magnetic fields, respectively, that exert external forces to separate BMSCs. However, there is no evidence assessing their influence on MSC functions and therapeutic effects (43, 44). Therefore, the crucial aspect of this method is specific surface markers. ISCT published minimal guidelines for the isolation of human BMSCs based on the positive markers CD105, CD73, and CD90 and the negative markers CD45, CD34, HLA-DR, CD79a, CD19, CD11b, and CD14 (21), and we summarized a variety of unique markers for human BMSCs in Table 1 (45–58).

The scientific studies and clinical applications of BMSCs require substantial expansion ex vivo to obtain sufficient numbers of cells because MSCs are rare in the bone marrow (0.001-0.01% of total nucleated cells and 0.42% of plastic adherent cells) (19). To maintain the functions and activities of BMSCs in vitro, a variety of approaches have been used to optimize the culture conditions, including medium composition, cell seeding density, passages and parameters related to the external environment, such as oxygen tension, pH and temperature.

MCSs can inhibit T‐cell proliferation and activation and induce the differentiation of Tregs. Several soluble immunosuppressive factors secreted by MSCs are involved in T-cell immunoregulation; the release of HLA-G1, TGF-β and HGF induces cell cycle arrest in G1 phase by downregulating phosphoretinoblastoma (pRb), cyclin D and cyclin A as well as upregulating cyclin-dependent kinase inhibitor 1B (p27Kip1), rendering T-cell activation ineffective (113, 114), and the production of IDO after IFN‐γ stimulation promotes tryptophan metabolism, resulting in the depletion of tryptophan, which inhibits proliferation and induces apoptosis of T cells (115). The direct interaction between MSCs and T cells may trigger T-cell apoptosis through the Fas ligand (FasL)-dependent pathway (116) as well as the PD-L1 pathway (117); of note, Fas ligand-associated T-cell apoptosis can induce macrophages to produce TGF-β, thereby increasing the abundance of Tregs (116). Additionally, several studies revealed that MSCs could inhibit the differentiation of Th17 cells through different mechanisms; the expression of CD54 recruited Th17 cells to MSCs and then upregulated PD‐1, IL‐10 and PGE2, blocking differentiation (118, 119). Some reports have also shown that activated T cells can differentiate into Th2 (120), CD3⁺CD45RO⁺ memory Treg cells (121), CD8⁺CD28^‐ Treg (122), IL10⁺ Tr1 and TGFβ⁺ Th3 (123) cells in the presence of MSCs, which suppress immune responses and accelerate tissue repair. Although the detailed mechanism of the interaction between MSCs and B cells is controversial, it is known that the inhibition of B-cell proliferation by MSCs seems to be associated with cell cycle disruption at specific stages rather than the induction of apoptosis (124, 125). Activated MSCs can interfere with the formation of plasmacytes and promote the differentiation of Bregs; for example, the soluble molecule IDO secreted by MSCs is involved in the survival and proliferation of CD5⁺ regulatory B cells (126), and when MSCs express IL-10, they promote the production of CD19⁺CD24⁺CD38⁺ Bregs in humans and CD19⁺CD1d high CD5⁺ Bregs in mice (126, 127). In addition, MSCs can promote the formation of naive, transitional and memory B-cell subsets, and these nonactivated B cells can induce Treg differentiation. Notably, the MSC-derived CC chemokine ligands CCL2 and CCL7 can suppress immunoglobulin (such as IgA and IgM IgG) production and release by plasmacytes (128).

Natural killer cells (NK cells) are important effector cells of innate immunity (129). MSCs can inhibit the proliferation, cytotoxic activity and cytokine production of resting NK cells (130). IL-2-mediated proliferation of resting NK cells is inhibited by coculture with MSCs. The mechanism may involve the release of IDO, PGE2 and HLA-G5 (131–133), and upregulating the expression of HLA class I molecules (MHCI) inhibits cytokine-mediated induced NK-cell cytotoxicity and decreases the secretion of cytokines (130). MSCs also maintain the activity of neutrophils for a long period to promote the elimination of invading bacteria (134), and the MSC-derived soluble factor IL-6 can delay apoptosis of neutrophils and inhibit the neutrophil-mediated fulminant inflammatory response, which is called the respiratory burst (135). The immunoregulatory effect of MSCs on macrophages mainly converts polarized inflammatory macrophages (M1) to anti-inflammatory macrophages (M2). The mechanism may involve IDO, TSG-6 and PGE2 (134, 136–138). Furthermore, proinflammatory factors (IFN‐γ, TNF‐α and LPS) can enhance the M2 macrophage polarization of MSCs. To date, the membrane-bound molecules CD54 and CD200 have been found to increase the immunosuppressive function of MSCs (139, 140). Regarding myeloid dendritic cells (mDCs), MSCs can inhibit the development and maturation of mesenchymal/dermal DCs and the conversion of umbilical cord blood and CD34⁺ haematopoietic progenitors as well as monocytes into DCs (141) (142, 143). Several recent studies suggested that MSC-derived IL-6, macrophage colony-stimulating factor (M-CSF), TSG-6 and PGE2 could be responsible for the immunoregulatory interaction between MSCs and immature dendritic cells (141, 142, 144). Furthermore, MSCs can induce the transformation of mature dendritic cells into immunosuppressive regulatory dendritic cells via jagged-2 and IL-10-activated SOCS3 pathways while escaping from their apoptotic fate (145, 146), and IL-10-plasmacytoid dendritic cells (pDCs) can be induced by PGE2 (124).

These soluble and membrane molecules play important roles in MSC immunomodulation, and researchers can evaluate the effectiveness of pretreatment in vitro for improving the immunomodulation of MSCs. Recently, the concept of immune training of MSCs has been proposed (147–149), where MSCs are stimulated in vitro by proinflammatory factors or cocultured with activated immune cells, and when MSCs are stimulated again with the same stimuli, detection of the immunosuppressive molecules we mentioned can be used to assess whether MSCs can achieve “memory” to rapidly and efficiently suppress inflammatory signaling.

Devascularization models mainly imitate hepatic ischaemia-reperfusion injury (IRI) caused by liver transplantation, hepatectomy and haemorrhagic shock (150, 151), and such models have also been used to investigate liver regeneration and the therapeutic potential of artificial liver support systems (ALSSs) (152, 153). Hepatocytes and endothelial cells experienced hypoxic insult during a brief period of ischaemia. Subsequently, dysfunctional mitochondrial respiratory chain-activating degradative enzymes cause a range of disruptions in intracellular proteins, lipids and DNA. Reperfusion produces reactive oxygen species (ROS) and hydroxyl radicals that activate Kupffer cell amplification cascades and inflammatory responses, recruit neutrophils (154), and trigger different types of cell death, including apoptosis, autophagy-associated cell death and necrosis (155).

As many investigations have shown, ischaemia-reperfusion models in rats (156), mice (157) (C57BL/6 mice were described as the most popular model) and pigs (158) (in combination with hepatectomy) have been developed via 70% hepatic segmental thermal ischaemia. The critical protocol is a noninvasive vascular clip on the upper left side of the portal triad structure (bile duct, portal vein and hepatic artery) for 45 (159), 60 or 90 (160) minutes to block the blood supply to the left and median lobe of the liver, and reperfusion is initiated by removal of the clamp (161). A previous study revealed reproducible hepatic injury at 60 min of ischaemia and is therefore extensively employed in ischaemia-reperfusion models (157, 161–164), allowing decompression of the portal vein through the right lobe and caudate lobe. To prevent mesenteric vein congestion, all surgical procedures were performed at a constant temperature of 37°C.

Tail vein (165, 166), hepatic vein (167) and peripheral vein (168) injection of MSCs in an ischaemia-reperfusion injury animal model showed that MSC infusion can reduce liver damage and cell death (165, 167), improve the levels of ALT and AST (166–168) and mainly decrease oxidative stress caused by liver excision and ischaemia-reperfusion injury (169). MSCs also inhibit the production of proinflammatory cytokines (TNF-α, IL-1β and iNOS), macrophage activation and neutrophil recruitment and promote anti-inflammatory cytokine secretion (IL-10), which is beneficial for recovery from liver injury and inflammatory responses (167). The potential mechanism of MSCs in the treatment of liver IRI may increase CD47 expression in the liver, and then the CD47-SIRPα interaction activates HEDGEHOG/SMO/Gli1 signaling and further inhibits NEK7/NLRP3 activity to protect the integrity of the liver (166). Zheng et al. found that MSCs upregulated PINK1-dependent mitophagy and exerted a protective effect in liver IRI, which might be associated with the modulation of AMPKα activation (168).

Acetaminophen-induced drug-induced liver injury (DILI) is the most frequent cause of ALF in many Western countries, such as the United States and the United Kingdom. The animal models caused by acetaminophen are more similar to the pathophysiological characteristics of liver failure in humans (170). The toxicity mechanism of excess acetaminophen-induced oncotic necrosis begins with the accumulation of a toxic metabolite, N-acetyl-benzoquinone imine (NAPQI), catalysed by the cytochrome P450 enzyme system (171). Subsequently, a large amount of reactive oxygen species are formed, which initiates severe mitochondrial oxidative stress (172), mainly through activation of MAP kinase and translocation of phospho-c Jun N-terminal kinase (p-JNK) to mitochondria (173); the mitochondria eventually undergo swelling of the cytoplasm and rupture of the outer membranes (174), with the release of endonucleases to degrade nuclear DNA (175, 176). Several species, such as mice, rats, rabbits, dogs and pigs, have been used to develop acetaminophen models, and mice are widely used because they develop liver failure very close to those reported in humans in pathophysiology structure and acetaminophen doses (177).

Furthermore, the acetaminophen dose, the diluent, the administration route and the mouse strain are critical factors that need to be considered for model development. Many studies have noted that because glutathione can relieve the toxicity associated with NAPQI (178), fasting before acetaminophen administration maintains baseline levels of glutathione in all animals, which increases the consistency of experimental results and the success of acetaminophen-induced liver failure. In previous studies, typical hepatotoxicity was observed in fasted mice at doses of 200-300 mg/kg and in nonfasted mice at doses of 500-600 mg/kg or higher, and acetaminophen was diluted in normal saline (NS) or phosphate-buffered saline (PBS) and administered intraperitoneally to mice, while intravenous or subcutaneous administration is more suitable for large animals.

Some studies have reported that tail vein transplantation of MSCs significantly improves the survival rate of mice with liver failure induced by APAP and ameliorates liver function by reducing intense centrilobular necrosis and inflammatory infiltration (179–183). A recent study showed that MSC therapy can efficiently improve APAP-induced mitochondrial dysfunction and liver injury by inhibiting c-Jun N-terminal kinase (JNK)-mediated mitochondrial retrograde pathways (179). Another study reported that MSC-mediated immunoregulation is associated with the activation of the Notch2/COX2/AMPK/SIRT1 pathway (183). More interestingly, MSCs can enhance antioxidant activity to attenuate liver damage by inhibiting cytochrome P450 activity (by reducing NAPQI production) to reduce the depletion of GSH and oxidative stress. These results might be related to the downregulation of MAPK signalling and the decreased inflammatory responses (180).

Carbon tetrachloride (CCL4), a classical hepatotoxin, induces DILI after single high-dose administration and can progress to chronic liver disease (CLD) (184) such as nonalcoholic steatohepatitis (NASH) (185), hepatocellular carcinoma (HCC) (186), acute-on-chronic liver failure or other alcohol-related liver disease and fatty liver disease after multiple low-dose administrations (187). After entering the body, CCL4 depends on the cytochrome P450 enzyme system metabolism for conversion into reactive trichloromethyl radicals with high activity (188). These metabolites can cause lipid peroxidation and hepatocyte membrane rupture, as well as DNA strand breakage. In addition, it has been revealed that such damage further affects the transcriptional and replication activity of hepatocytes, resulting in portacaval zone necrosis (189).

In general, CCL4 administration is performed in mice, rats and rabbits (BALB/c mice have been described as the most appropriate model) via intragastric administration, intraperitoneal injection, subcutaneous injection or inhalation to induce acute or chronic liver failure. In some investigations, 6- to 8-week-old male mice (weighing approximately 25-30 g) were used to develop acute mouse models of CCL4 induction (190, 191), in which olive or corn oil served as diluents to solubilize CCL4 at ratios ranging from 10% (v/v) to 50% (v/v), and CCL4 doses of 2 mL/kg or higher can induce acute liver failure in mice (191–194).

MSC therapy efficiently prolonged the survival time of CCL4 induced acute liver failure mice from day 2 to day 7 after transplantations of second trimester amniotic fluid (AF-MSCs), and the ALT and AST levels significantly decreased by 35.36% and 64.72%, respectively (195). In addition, Milosavljevic et al. found that MSCs can modulate the IL17 signaling to treat the immune-mediated liver failure via altering NKT17/NKTreg ratio and suppressing hepatotoxicity of NKT cells in an IDO-dependent manner (196). A previous study reported that the treatment effect of adipose tissue-derived mesenchymal stem cells (AT-MSCs) may relate to the secretion of interleukin 1 receptor (IL-1R), IL-6, IL-8, granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), monocyte chemotactic protein 1, nerve growth factor, and hepatocyte growth factor (197).

Lipopolysaccharide (LPS) is a molecule that is present in the outer membrane of gram-negative bacteria and can activate Kupffer cells (198–200), triggering the secretion of multiple inflammatory mediators (200, 201), and the coadministration of 300 mg/kg D-Gal dramatically increased rodent susceptibility to LPS, resulting in extensive liver injury and cell death (202). D-Gal is a hepatocellular phosphate uracil nucleotide interference agent that is metabolized via the galactose pathway and can cause diffuse necrosis and inflammation rather than zonal necrosis, similar to most hepatotoxic drugs (153). The administration of D-Gal/LPS in mice induced liver necrosis and inflammation similar to human hepatitis (203, 204). Previous studies demonstrated that coadministration of LPS at doses of 10-100 µg/kg and D-Gal at doses of 100-1000 mg/kg was performed to establish an acute liver failure mouse model (205–209).

The pathophysiological mechanism of D-Gal/LPS-induced ALF involves the binding of LPS to Toll-like receptor 4 (TLR4) on Kupffer cells, which triggers transcriptional and translational activation of cytokines such as TNF-α, IL-1β and IL-6 (210, 211). In particular, TNF-α has been recognized as a key regulator of hepatitis, as it recruits many neutrophils into the liver sinusoids and induces the expression of various adhesion molecules, including intercellular adhesion molecule 1 (ICAM1), vascular adhesion molecule 1 (VCAM1), selectin, and chemokines on endothelial cells and hepatocytes, after LPS treatment (212, 213). Some of these adhesion molecules are critical for neutrophil extravasation and cytotoxicity. In addition, by binding to its receptor 1 (TNFR1) on hepatocytes, TNF-α activates the nuclear factor kappa beta (NF-κB) pathway, resulting in the expression of proinflammatory and antiapoptotic genes (214, 215). Although high doses of D-gal inhibite the synthesis of antiapoptotic genes by depleting uridine triphosphate in hepatocytes, they promote apoptosis signaling via activation of the caspase cascade and DNA damage. Thus, TNF-α-mediated apoptotic signaling and inflammation are commonly considered pathophysiological mechanisms of D-Gal/LPS-induced ALF. Notably, the interaction between these mechanisms remains uncertain and should be explored in the future.

BMSC transplantation rescued the D-gal-induced liver failure model. In rodents, the 4-week survival rate significantly increased by 80% (216). Numerous hepatocytes were repaired, with only a few necrotic areas. In D-gal-induced liver failure in large animals, BMSC therapy significantly prolonged the survival time from 3.22 days to more than 14 days by suppressing the life-threatening cytokine storm. BMSC-derived hepatocytes were widely distributed in injured livers within 10 weeks, with liver function returning to normal levels (217). During recovery, serum levels of proinflammatory molecules, including IFN-γ, IL-1β, and IL-6, were reduced, while serum levels of the anti-inflammatory cytokine IL-10 were significantly increased through paracrine effects, referring to regulation by the STAT3 signaling pathway (216) and notch-DLL4 signaling pathway (218).

Fas receptor (CD95), a member of the TNF-receptor superfamily with a death domain, mediates the assembly of a death-inducing signaling complex. Inducing caspase activation and cell apoptosis (219) has been considered the critical mechanism of fulminant liver failure (220), ischaemia-reperfusion-associated liver diseases (221), nonalcoholic fatty liver disorders (222) and other acute and chronic hepatic disorders. The liver constitutively and abundantly expresses the Fas receptor and activated caspase (casp) 8 upon binding of FasL or other receptor agonists (220, 223), such as the Fas receptor antibody JO2 and soluble FasL in the hexameric form (MegaFasL) (198). Then, it triggers the caspase cascade accompanied by excessive hepatocyte apoptosis (224), which can quickly progress to secondary necrosis (225). This mechanism may rely on activation of Fas-induced inflammatory signaling via the nonclassical interleukin-1β pathway (225).

Several researchers have reported that JO2 at concentrations such as 0.15, 0.2, 0.23, 0.35, 0.4, 0.42, and 0.5 mg/kg can induce liver failure, and the severity of liver injury is dependent on the JO2 dose (220, 221, 225–231). Thus, the critical element for developing this animal model is the concentration of JO2. Shao et al. (220) observed severe liver damage, including destruction of the hepatic lobules, hepatocyte necrosis and haemorrhage after treatment with 0.5 mg/kg JO2 (dissolved in normal saline (NS)) in BALB/c mice via intraperitoneal injection. However, after treatment of C57BL/6 mice with the same doses and methods, all mice died within 12 h (231). Although this difference could be caused by differences across researchers or other environmental factors, strain differences. In addition, the administration route can affect the pharmacodynamics and pharmacokinetics of the drug.

In recent works, BMSC transplantation rescued mice with JO2-induced liver failure and prolonged the survival time by improving liver function and decreasing extensive hepatic necrosis and haemorrhage. BMSCs can colonize injured mice and transdifferentiate into hepatocytes and cholangiocytes, and many KRT7- and KRT19-positive human cholangiocytes form tubular structures around the portal area (22). Meanwhile, transplanted BMSCs differentiate into immune cell lineages, including T cells, B cells, natural killer (NK) cells, macrophages and dendritic cells, and play a paracrine role by regulating inflammatory cytokine levels (23). Another study confirmed this finding and further identified the two transdifferentiation phases by transcriptomics; hepatic metabolism and liver regeneration were characterized in the first 5 days after BMSC transplantation, and immune cell growth and extracellular matrix (ECM) regulation were observed from day 5 to day 14 (25).