The β-catenin-dependent canonical WNT pathway appears to be the most conserved pathway in both vertebrates and invertebrates and plays a dual role in adherent junctions and transcriptional regulation (37–39). The receptors in the WNT signaling pathway include single-pass transmembrane co-receptor LRP5/6 and seven transmembrane signaling receptors, FZD (44). The key element is the destruction complex consisting of the scaffold protein Axis inhibitor (AXIN), adenomatous polyposis coli (APC) and casein kinase 1 (CK1). Without WNT ligands, the destruction complex in the cytoplasm is activated to phosphorylate downstream β-catenin. APC binds β-catenin and CK1 and primes the protein for the subsequent phosphorylation mediated by recruited glycogen synthase kinase 3 (GSK3) at threonine and serine residues. The phosphorylated β-catenin is then degenerated by E3 ubiquitin ligase and transferred to proteasomes (Fig. 1A). When the WNT ligands are bonded, dishevelled (DVL) protein is recruited by FZD, and the phosphorylated DVL then phosphorylates and inhibits GSK3. β-catenin is thus protected from GSK3-dependent phosphorylation. Subsequently, non-phosphorylated β-catenin is translocated into the nucleus, where it displaces the co-repressor groucho from the transcriptional factor groucho/T cell factor/lymphoid-enhancing factor (TCF/LEF) complex, thereby triggering the transcription of WNT target genes, which mediate MSC bioactivity (Fig. 1B) (45–47).

To date, two distinct noncanonical WNT signaling pathways have been identified, with different impacts on cell activities. Similar to canonical WNT signaling, DVL is recruited and activated, and the subsequent cascade is meditated by the DVL-associated activator of morphogenesis 1 or DVL-associated small signaling G-protein Rac1. One of these pathways is the WNT-PCP pathway, which is initiated following binding of WNT to FZD together with a transmembrane co-receptor, followed by the activation of the small G-protein Rho by either orphan receptor tyrosine kinase-like receptor 2 (ROR2) or receptor tyrosine kinase through a guanine exchange factor (GEF), followed by the activation of Rho-associated protein kinase (48), an important regulator of the cytoskeleton. As the alternative pathway, Rac1 activates transcription factors through cJNK activation, which is involved in cytoskeleton modification, as well as targets the expression of genes regulating cell survival, movement and polarity (Fig. 2A) (49,50).

Another signaling pathway is the WNT-calcium pathway, which reduces cell adhesion (Fig. 2B). Phosphorylated DVL, following the binding of WNT ligands to FZD, leads to phospholipase C activation and the subsequent formation of inositol-1,4,5-triphosphate (IP3) and diacylglycerol (DAG) from the cell membrane component, phosphatidylinositol-4,5-bisphosphate (PIP2). IP3 binds to its receptor in the endoplasmic reticulum (ER) membrane, resulting in the release of intracellular calcium (47,49,50). Calcium movement in the cytoplasm then activates calcium-/calmodulin-dependent protein kinase II (CAMKII), which causes nuclear translocation of activated T cells (NFAT) family of transcription factors through the phosphorylation of MAPKs (51), while DAG activates PKC and MAPKs, which all target NFAT inside the nucleus to govern cell movement and adhesion. In addition, the noncanonical WNT-calcium pathway has been reported to inhibit the canonical WNT/β-catenin pathway, since CAMKII can trigger TGF-β activated kinase 1 (TAK1) phosphorylation, which in-turn enhances the activity of Nemo-like kinase (NLK), leading to LEF1-β-catenin/DNA dissociation (Fig. 2B) (52). Recently, another special noncanonical pathway leading to epithelial-mesenchymal transition in cancer cells was discovered (53). Upon WNT5a ligand binding, FZD2 exclusively recruits and phosphorylates DVL. The pathway is achieved through the activation of STAT3 by FYN. STAT3 is translocated to the nucleus and triggers cell migration-related target gene expression, epithelial-mesenchymal transition and tumor cell invasion responsible for tumor metastasis (Fig. 2C) (53).

Overall, the WNT signaling pathway inhibits MSC adipogenic differentiation. It has been demonstrated that the expression levels of WNT antagonists sFRP4 and dickkopf WNT signaling pathway inhibitor 1 (Dkk1) were higher in adipogenically differentiated than in undifferentiated MSCs (59). Moreover, a short 48 h treatment of human MSCs with secreted frizzled-related protein (sFRP)1 and sFRP4 was reported to upregulate adiponectin secretion, thereby promoting adipogenesis (60,47). A few clinical trials explored the association between WNT activity, and obesity and diabetes. One of these trials concluded that sFRP4 levels were positively correlated with impaired glucose and triglyceride metabolism (61). Certain studies have used sFRP4 as a predictor of type II diabetes mellitus (62,63). Conversely, a previous study showed that the GSK-3β inhibitor 6-bromo-indirubin-3α-oxime could inhibit MSC adipogenesis, as a result of enhanced WNT signaling (64). Collectively, this evidence showed the inhibitory regulation of WNT on the differentiation of MSCs into adipocytes, suggesting that WNT activators can be used for the treatment of obesity, although this remains under extensive investigation.

There are a number of controversies regarding the function of the WNT pathway in MSC chondrogenesis, which may be a result of the complicated interactions within the WNT signaling network. It was previously shown that an enhanced chondrogenic activity on sFRP1-deficient mice highlighted the positive effect of WNT signaling on MSC chondrogenic differentiation (65). In addition, the activation of noncanonical WNT following WNT5a ligand binding was also found to induce the chondrogenesis of MSCs derived from the chicken wing bud (66). Similarly, an in vitro study showed that WNT3a overexpression was associated with the chondrogenic differentiation of C3H10T1/2 murine mesenchymal cells (67). The increase in pre-cartilage condensation in an ex vivo MSC culture in the presence of WNT5a protein further confirmed a promoting effect of WNT signaling on MSC chondrogenesis (66). The underlying mechanism indicates that WNT5a activates the noncanonical WNT pathway and blocks the canonical pathway. However, a different study reported a reduced expression of chondrogenic-specific markers, such as collagen II, SRY-box transcription factor (Sox)9 and aggrecan following continuous treatment with WNT1 protein for 21 days, whereas treatment with the WNT antagonist Dkk1 increased their expression in human adipose-derived MSCs (68). Consistently, a large number of in vitro studies emphasized the importance of suppressed WNT signaling during chondrogenesis. For example, sFRP1 and Dkk1 could accelerate the initial stages of human MSC chondrogenesis, as indicated by the enhanced glycosaminoglycan synthesis, Sox9 and type II collagen expression (69); these WNT antagonists also exhibited a chondrogenesis-promoting effect in long-term pellet cultures (70). Most of these conclusions were made based on in vitro studies, where the concentration of the WNT pathway regulators and the absence of other interacting pathways could be the reason for the controversial findings. Therefore, these findings warrant additional and more accurate investigations.

It has been well-established that the WNT signaling pathway is indispensable for MSC osteogenesis. The canonical WNT signaling pathway activates the differentiation of MSCs into osteoblasts (71). Current research also suggests that the canonical WNT pathway is essential for the contribution of Gli1+ MSCs to alveolar bone development and the blocking of the canonical WNT pathway in catenin conditional knockout mice, which caused severe bone loss (72). The disruption of WNT signaling by a functional mutation or targeted destruction of LRP5 in mice has been shown to promote osteoporosis and a low bone mass phenotype (73,74), whereas its overexpression has been shown to lead to a high bone mass syndrome (75). The WNT/β-catenin signaling functions like a switch, favoring MSC osteogenesis at the expense of adipogenesis by modulating the availability of cell-type-specific transcription factors (76). Recently, Src homology region 2 domain-containing phosphatase-1 (SHP1) was found to bind with GSK3 and suppress its kinase activity. In SHP1 partial deficient mice (mev/mev), phosphorylated GSK3 mediated increased β-catenin degradation, thus these mice developed osteoporosis. Lineage differentiation culture of the bone marrow MSCs extracted from mev/mev mice displayed less osteogenesis and more adipogenesis (77). As expected, WNT antagonism was reported to inhibit osteogenesis. This inhibitory effect was demonstrated by an in vitro study, where sFRP4 inhibited the periodontal MSCs committed to osteogenic progenitor cells (78). Amongst other WNT antagonists, sFRP1 overexpression was found to inhibit in vivo bone formation (79) and deteriorate osteoblast and osteocyte apoptosis (80). Comparatively, the lack of sFRP1 reduced apoptosis, accelerated MSC osteogenic differentiation (81), enhanced trabecular bone formation and improved fracture healing (82–84), confirming the inhibitory effect of WNT antagonists on osteogenesis. In addition, sFRP3 has been shown to repress the noncanonical WNT pathway by binding to WNT5a, which in turn blocks the inhibitory effect of WNT5a on the canonical pathway, hence promoting osteogenesis (78). Collectively, this evidence supported the positive regulation of the canonical WNT pathway on MSC osteogenic differentiation. The extensive crosstalk within the WNT signaling network during osteogenic differentiation should be considered when devising possible new therapeutic approaches for bone-related diseases.

TLRs are dominantly expressed on the membranes of leukocytes, including dendritic cells, macrophages, natural killer cells, and T and B lymphocytes. To date, 13 members of the TLR family have been identified, of which TLR1, TLR2, TLR4, TLR5, TLR6 and TLR11 are expressed on the cell surface, and TLR3, TLR7, TLR8 and TLR9 are localized to the endosomal/lysosomal compartments. TLR12 and TLR13 are not found in humans (106). Each of these TLRs binds to and becomes activated by different ligands. For example, lipopolysaccharides (LPSs) exclusively bind to TLR4, but TLR4 can also be activated by multiple other pathogens. TLR3 can be activated by Poly I:C-a mismatched double-stranded RNA in viral infection (107). Lipoteichoic acid (108), the immunostimulant of Gram-positive bacteria, is recognized by TLR2. The inflammatory pathways initiated after this receptor-ligand binding and their downstream responsive adaptors are also different for each TLR. However, overall, all the TLR inflammatory pathways act through two major signaling cascades: A MyD88-dependent pathway inducing pro-inflammatory cytokines, such as TNF-α via NF-κB and a MyD88 independent pathway that acts through type I interferons (109). The dominant MyD88-dependent pathway originates from the cytoplasmic Toll/IL-1 receptor (TIR) domain. Upon stimulation with extracellular PAMPs, MyD88 attaches to TLRs through the TIR domain and recruits IL-1 receptor-associated kinase-4 (IRAK-4) to TLRs. Following phosphorylation, the IRAK1/2/4 complex is connected to TNF receptor-associated factor 6 (TRAF6), which then activates TAK1, leading to the activation of the IκB kinase (IKK) complex. The downstream IκB kinase phosphorylates the inhibitory IκBα protein (73,110). This phosphorylation dissociates IκBα from NF-κB, with the free NF-κB then able to translocate into the nucleus to activate the target genes responsible for pro-inflammatory cytokine transcription, such as pro-IL-1β and pro-IL-18. The enhancement of the transcription of the same target genes can also be mediated by activated mitogen-activated protein kinase kinases to phosphorylate MAPKs including JNK, ERK and p38 (111). The activation of MyD88-independent pathways occurs via TIR-domain-containing adapter-inducing interferon-β and TRAF3 in case of MyD88 deficiency. It induces the recruitment of IKKε/TBK1, phosphorylation of interferon regulatory factor 3 (IRF3) and expression of interferon-β. Thus far, most TLR-related inflammation is mediated by the MyD88-dependent pathway, whereas PolyI: C-TLR3 antiviral innate immunity has been found to be independent of MyD88 (Fig. 3A) (112).

TLRs and IL-1R ligands are usually found to be highly expressed in sites of tissue injury and wound repair (113,114), and it has been shown that they could influence the repair process (115,116). Thus, it is reasonable to conclude that TLR inflammatory pathways have an impact on the behavior of MSCs. It has been well established by several in vitro studies that MSCs express a number of TLRs. Thus far, consistent results have demonstrated the expression of TLR1-6 in adipose- and bone marrow-MSCs in both humans and mice and at both the mRNA and protein levels, whereas inconsistent results have been reported on the expression of TLR7-10 (84,117). TLR/Myd88/IL receptor 1 (ILR1) signaling inhibits BM-MSC colony formation, proliferation, migration and osteoblastic differentiation after being activated by IL-1β. It is worth mentioning that BM-MSCs are more vulnerable to the inhibitory effects of TLR/Myd88/ILR1 signaling than osteoblasts (30). Pevsner-Fischer et al (118) found that TLR2 activation reduced mouse BM-MSC differentiation into the three mesodermal lineages. Of note, BM-MSCs collected from MyD88-deficient mice were found to effectively differentiate into adipocytes, but not osteocytes or chondrocytes, even without the additional stimulation with TLR ligands. Similarly, another two studies reported that the activation of the TLR4 or 2 pathway in the proinflammatory environment negatively regulated human adipose MSC (hAD-MSC) differentiation to adipocytes (119,120). Collectively, these findings indicated that the effects of the TLR pathway on MSC lineage commitment potentials varies depending on different culture conditions, tissue origins and species (118). However, most of the available evidence on the association between the infectious TLR pathway and MSCs are based on in vitro studies. In vitro cultures do not completely replicate in vivo MSC behavior, as it is unable to incorporate the complicated interactions between the numerous signaling pathways employed by the various cell types. A good example of this difference between in vivo and in vitro studies is the fact that in BM-MSC-based calvaria regeneration, delivering Myd88^−/−MSCs induced a significantly higher regeneration capacity compared with wild-type (wt) MSCs, whereas the in vitro culture of Myd88^−/− and wt MSCs resulted in equal differentiation and proliferation ability (30).

It is well known that the oral cavity is subjected to daily assaults from the external microbiota, which inevitably causes periodontitis. In periodontitis, LPS can cause periodontal tissue degeneration in a dose-time-dependent manner (121), which indicates the failure of periodontal MSCs to support physiological turnover. The modulation of WNT/β-catenin signaling helps attenuate periapical bone lesions (122). An in vitro study demonstrated that LPS activated p38-MAPK and inhibited the canonical WNT/β-catenin signaling pathway with an increase in the levels of p38, c-myc, cyclin D1 mRNA and phosphorylated GSK-3β (123). During the process of wound repair, prostaglandin E2 (PGE2) released from neighboring apoptotic cells can activate the WNT/β-catenin pathway in MSCs via WNT3. LPS was also reported to enhance WNT5a expression through the TLR4/MyD88/phosphatidylinositol 3-OH kinase/Akt/NF-κB/MAPK pathway, based on in vitro studies of human dental pulp stem (124,125) and osteoblast (126) cells. However, there are a number of controversies on the influence of Pg-LPS on dental MSC function and the pathways involved. It has been shown that Pg-LPS modified periodontal ligament stem cell lineage commitment during the inhibition of osteogenesis and stimulation of fibrosis via ERK1/2 signaling (127). Moreover, Pg-LPS inhibited the osteogenic differentiation of BM-MSCs (127), whereas LPS upregulated gingival MSC proliferation without attenuating their regenerative capacity, and this positive effect was mediated by the NF-κB but not the WNT/β-catenin pathway (127).

The digestive system is another non-sterile environment that is exposed to a wide array of microorganisms, with infectious bowel diseases being relatively common as a result (128). Under infectious conditions, intestinal MSCs can differentiate into inflammatory cells and secrete inflammatory cytokines, such as IL-1b, IL-8 and TNF-α (129). When treating intestinal MSCs with LPS, flow cytometry analysis showed that the LPS-induced MSC cell cycle progression was arrested at the G1 phase, and that cell pyroptosis was enhanced. Studies have shown that LPS inhibits WNT signaling through the activation of GSK3β, causing a marked inhibition in enterocyte proliferation, both in vitro and in vivo (130). This particular phenomenon links the LPS/TLR4 pathway to the regulation of intestinal MSC lineage commitment by WNT.

Of note, Liu et al (131) found that although Salmonella effector Protein AvrA presents in certain intestinal microorganisms of Salmonella and E. coli, it specifically upregulates the expression of WNT genes and phosphorylation of β-catenin, thus leading to a marked increase in crypt proliferation with a significant reduction in Lgr5+ intestinal stem cells in their model (131,132).

The above findings suggested an influence of the TLR inflammatory pathway on the regulatory WNT pathway in MSCs, with the molecular mechanism of this interaction partially revealed. TLRs can modulate WNT activity at different points with various effects. First, TLRs can modulate the activity of WNT degradation of β-catenin by binding to the LRP5/6 FZD receptor complex, thus promoting the function of the β-catenin destruction complex, with subsequent blocking of the WNT target gene expression. However, this interactive mechanism helps explain the fact that transiting WNT3a could counteract the inhibitory effect of Pg-LPS. TAK1 phosphorylation in TLR signaling attributes to NLK activation, which is negatively correlated with the canonical WNT pathway downstream TCF/LEF transcription (133). By contrast, TLRs can activate AKT through PI3K and IKK, consequently inhibiting GSK3β, enhancing β-catenin expression and upregulating TCF/LEF transcription. TLR4/Myd88/leucine-rich repeat (LRR) binding FLII interacting protein 2 has also been revealed as an activator of WNT by interacting with DVL to activate catenin/LEF/TCF-dependent transcriptional activity (Fig. 3A and B) (134,135).

In addition to TLRs, vertebrates have developed several alternative strategies to sense pathogens in the cytosol (136,137). NLRs are the intracellular receptors playing key roles in innate immune system regulation by sensing PAMPs that enter the cell via phagocytosis or through pores, and cell stress-related damage-associated molecular patterns (106). NLRs are composed of 3 major domains: The central NOD nucleotide-binding domain (NACHT) domain is common to all NLRs; most NLRs also have similar C-terminal LRR and variable N-terminal interaction domains (114). Each domain has a specific function; LRR recognizes ligands, the central NACHT domain mediates adenosine triphosphate (ATP)-dependent self-oligomerization, and N-terminal domain triggers homotypic protein interaction amongst the caspase recruitment domain (CARD; from the NLRC subfamily), pyrin domain (PYD; from the NLRP subfamily), acidic transactivating domain (from the NLRA subfamily) or baculovirus inhibitor repeats (BIRs; from the NLRB subfamily) (138,139).

NLRs mediate inflammasome formation and downstream caspase-1 activation, resulting in programmed cell death under infectious conditions, a phenomenon known as pyroptosis. The manifestation of pyroptosis can be induced by pyknosis, chromatin condensation, DNA breaks, plasma membrane permeabilization and cellular swelling (140). Following TLR-related ‘priming̛ and ‘activating̛, NLR ligand uptake triggers inflammasome assembly, and maturation of IL-1β and IL-18. NLRs directly bind to apoptosis-associated speck-like protein (ASC) through PYD, and ASC binds to pro-caspase-1 through its CARD domain (141). NLR signaling-driven pyroptosis can be mediated by the canonical and noncanonical inflammasome pathway. In the canonical inflammasome pathway, the combination of NLRs/ASC activates pro inflammasome 1 and organizes the functional caspase-1 into a multiprotein oligomer complex. The activated caspase-1 induces IL-1β and IL-18 secretion from pro-IL-1β and pro-IL-18. As a consequence, inflammatory immune cells are attracted, and pro-inflammatory cytokines (such as TNF-α and IL-6), chemokines (such as IL-8, CXCL8 and monocyte chemoattractant protein 1) and adhesion molecules are released. In addition, the K⁺ selective hemichannel on the inflammatory cell surface is opened upon ATP binding, with the subsequent K⁺ efflux inducing pore formation on the cell membrane via pannexin-1, and extracellular pathogen influx further deteriorating pyroptosis. The noncanonical pathway works in synergy with the canonical pathway. In the non-canonical pathway, caspase-11 activation directly enhances IL-1β and IL-18 release and promotes the canonical caspase-1 pathway (142). Subsequently, the release of IL-1β and IL-18, and proteolytic cleavage of Gasdermin-D into Gasdermin-N domain by caspase-4, 5 and 11 either directly activate caspase-1 or indirectly activate the canonical inflammasome/NLR pathway to induce pyroptosis (142,143).

A previous in vitro study on human umbilical cord blood-derived MSCs (hUCB-MSCs) and hAD-MSCs reported that hUCB-MSCs expressed bioactive NOD1 and 2 genes. The administration of NLR agonists has led to IL-8 production. Activated NLR pathways enhance hUCB-MSC osteogenesis and chondrogenesis, and inhibit adipogenic differentiation, but with no marked effects on MSC proliferation (144). Controversially, another in vitro BM-MSC culture study demonstrated increased MSC adipogenesis and decreased MSC osteogenesis in the presence of the NLRP3 inflammasome, which may be explained by the fact that these stem cells originated from different tissues. The inhibition of caspase-1 activation offered a novel therapeutic target for ageing-related chronic inflammatory diseases, such as osteoporosis (145). Despite the contrasting results, the aforementioned findings suggest that NLR pathways serve a pivotal role in MSC regulation. The literature discussing the impact of NLR pathways on WNT signaling is very limited, and the available evidence suggests that receptor-interacting protein kinase 2 (RIP2) containing the CARD domain binds to NLR-inflammasome and undergoes phosphorylation. The NLR pathway can negatively regulate the WNT pathway via the TAK/NLK cascade (146–148). Of note, certain studies have reported that NOD2 can function as a positive regulator of WNT signaling. The activation is achieved in a noncanonical, RIP2-TGFβ-TAK1-independent manner (149–151). As an alternative approach, NOD2 was found to stimulate and mediate the WNT pathway through Ly6/PLAUR domain-containing protein 6, which synergizes the ligand-receptor combination in the WNT pathway (Fig. 3C and D) (149).

Various in vivo and in vitro studies have shown that the inflammatory cytokines secreted during tissue infections may affect stem cell activities. In most cases, proinflammatory cytokines have been shown to repress various aspects of MSC behavior. The endogenous pro-inflammatory cytokine IL-1β activates IL-1R/MyD88 signaling, thus impairing MSC proliferation, migration and differentiation by inhibiting the AKT/GSK-3β/β-catenin pathway (30). Loss of function studies have shown that IL-1β-initiated IL-1R/Myd88 signaling impairs stem cell proliferation, migration and differentiation (152). IL-1β and IL-18 have also been identified to directly kill stem cells in vitro (153). Similarly, IL-1β was found to inhibit BMSC osteogenesis by activating Foxhead box D3 protein-mediated miR-496 expression (154). Interferon-α (IFN-α) acts as a central mediator of adaptive immunity, and has been shown to stimulate bone loss during inflammation (155). The combination of IFN-γ and TNF-α can accelerate MSC apoptosis through the internalization of Fas cell surface death receptor, with a reduction in the expression of anti-apoptotic factors, such as NF-κB, X-linked inhibitor of apoptosis protein and FLICE-like inhibitory protein, whereas TNF-α alone may induce MSC apoptosis in a dose-dependent manner. The administration of recombinant TNF-α has been shown to inhibit MSC adipogenesis and osteoblastogenesis, thus promoting bone resorption (156,157). IFN-γ and TNF-α secreted from lymphocytes may synergistically block MSC-based bone regeneration (93). Following treatment of bone marrow MSCs with IFN-γ alone, or a combination of IFN-γ and TNF-α in vitro, IFN-γ was found to block osteogenic differentiation at the same time as upregulating SMAD family member 6 and inhibiting Runt-related transcription factor 2 (158). Both IFN-γ and TNF-α were shown to promote the phosphorylation of p38 MAPK, which was associated with the downregulation of cyclooxygenase-2/PGE2, and this in turn may gradually reduce the immunomodulatory ability of MSCs (159).

The activation of TLR and NLR pathways may trigger the production of a common by-product, antimicrobial ROS (160). ROS are short-lived oxygen-containing molecules existing in the form of free redials, including superoxide anion (O₂⁻), hydroxyl radical (OH), hydroxyl ion (OH⁻) and nitric oxide (NO), as well as non-radical ROS, such as hydrogen peroxide (H₂O₂) (161). ROS production occurs under different physiological and pathological circumstances aided by active enzymes (162). ROS are ubiquitously found in the extracellular space (163), plasma membrane (164) and intracellular compartments [including mitochondria, where nicotinamide adenine dinucleotide phosphate (NADPH) oxidases are concentrated] (165), peroxisomes, ER (166) and cytosol (NO synthases and lipoxygenases) (167). However, amongst these, mitochondrial complexes I and III, and the corresponding NADPH oxidase isoform NADPH oxidase 4, are the major sources of ROS production that play pivotal roles in MSC regulatory niches (168). Several studies have concluded that local hypoxia is the culprit for the failure of MSC-based tissue regeneration and MSC ageing (169,170).

ROS influence MSC survival in a concentration-dependent manner. An appropriate level of ROS is necessary and advantageous to maintain cellular proliferation and survival; however, excessive ROS may cause cellular damage and dysfunction after initiating chemical reactions involving RNA, DNA, proteins and lipids (161,171,172). Eto et al (173) demonstrated that AD-MSCs were extremely sensitive to oxygen concentration, with only cells implanted <300 mm from an oxygen source surviving and the rest undergoing apoptosis. However, excess ROS has been confirmed to activate MAPK pathways and apoptotic proteins, and suppress antiapoptotic pathways (173,174). Consistently, antioxidants stimulate MSC proliferation (175).

ROS involvement in the regulation of MSC differentiation may stem from the difference in ROS levels within the differentiated and undifferentiated MSCs. During homeostasis, undifferentiated MSCs are predominantly supported by glycolytic energy with higher levels of glycolytic enzymes and increased lactate production. By contrast, MSC-differentiated osteoblasts rely more on oxidative mitochondrial metabolism (176). Therefore, MSCs exhibit relatively low levels of ROS, but low antioxidant activity may suggest that they are more sensitive to oxidative stress (177–179). In fact, the impact of ROS on MSC lineage differentiation significantly varies depending on the dosage. The in vivo and in vitro studies collectively suggested that excessive ROS could suppress osteogenesis but induce adipogenic differentiation (180). Highly well-defined levels of ROS may activate MSC differentiation into chondrocytes (144), adipocytes (181), osteocytes (46) and neurons (182) through the enhanced activity of regulatory signaling pathways, including the WNT pathway. A previous in vitro study verified that the exogenous addition of H₂O₂ inhibits TCF-mediated transcription and that β-catenin overexpression and use of antioxidants could rescue that inhibition. In addition, ROS concomitantly increases with ageing, with the aged MSCs reported to exhibit a reduced expression of WNT target genes, such as AXIN2 and TNF receptor superfamily member 11b, in 31-month-old mice compared with 4-month-old mice, thus diminishing osteogenesis (183). An in vivo study revealed that irradiation injury could cause intestinal tissue regeneration via the activation of WNT signaling, which was achieved through the ROS/hypoxia-inducible factor/WNT2b signaling axis (132). One explanation of this mechanism could be that superoxide-generating NADPH oxidase (Nox) induces ROS production, leading to the inactivation of nucleoredoxin (NRX) in the β-catenin destruction complex to activate the WNT/β-catenin pathway (184). Conversely, in the MSC niche, the binding of WNT ligand to its receptor complexes may trigger the sequential activation of Src kinase, Rac1-GEF-vav guanine nucleotide exchange factor 2 through Src-dependent tyrosine phosphorylation and Rac1. Activated Rac1 may in turn induce Nox1-derived ROS, which may result in the oxidation of NRX, with the oxidized NRX then detaching from DVL. Subsequently, liberated DVL suppresses the β-catenin destruction complex, resulting in the stabilization of β-catenin and the activation of WNT signaling (184,185).