NRF2 hyperactivation negatively regulated osteogenesis differentiation by inhibition of the RUNX2-dependent transcriptional activity in MC3T3-E1 cells. The levels of Runx2 mRNA did not change significantly during the process, and there is no direct interaction between NRF2 and RUNX2, which may be explained by the inclusion of other proteins between them. In addition, NRF2 can directly bind to the ARE-like sequence near OSE2 in the osteocalcin promoter to reduce the transcriptional activity of osteocalcin as well (27). NRF2 hyperactivation in primary osteoblasts from Keap1 knockout (Keap1^−/− ) mice leads to impaired osteogenic differentiation (28, 29). Nrf2 knockdown or knockout significantly enhanced osteoblast mineralization (30). The primary osteoblasts from Nrf2 knockout (Nrf2^−/− ) mice in two studies induces upregulation of the levels of the mRNA for Runx2, Alpl (alkaline phosphatase), Bglap (osteocalcin), Sp7 (osterix) or spp1 (osteopontin) (29, 30).

During oxidative stress conditions, upregulation of NRF2 may inhibit osteoblast differentiation. Low concentration of non-toxic H2O2 may inhibit differentiation and mineralization of MC3T3-E1 cells, which is accompanied by up-regulation of Nrf2 gene expression (31). N-acetylcysteine (NAC), a precursor of glutathione (GSH), mitigated the inhibition of osteoblast differentiation and mineralization by H₂O₂ or radiation-induced oxidative stress, accompanied by down-regulation of the NRF2/HO-1 pathway, however, these studies did not conduct rescue experiments to prove that the effects of NAC depended on the inhibition of NRF2 (32, 33). Some studies have shown that NAC may reduce oxidative stress level through non-NRF2 activation pathway, thus alleviating the damage of osteoblast differentiation, meanwhile, the activation of the NRF2 itself and NRF2-driven stress genes, were inhibited by NAC (34, 35). In addition, the downstream antioxidant enzyme of NRF2 may also affect osteoblastic differentiation. Heme oxygenase-1(HO-1), which is the downstream target protein of NRF2, can inhibit the maturation and mineralization of osteoblasts (32, 33, 36). Although NRF2 inhibition alleviates the radiation-induced osteogenic differentiation and mineralization damage, ZnPP IX, an inhibitor of HO-1, significantly restored radiation-induced osteogenic damage without inhibiting radiation-mediated activation of NRF2 (33). Quercetin alleviates H₂O₂-induced damage to osteoblast differentiation by downregulating HO-1 without effecting on NRF2 expression (37). These data suggest that the relationship between osteoblastogenesis and NRF2/HO-1 or ROS is rather complex ( Figure 2 ), and HO-1 may be a bigger culprit of oxidative stress-induced osteogenic injury.

NRF2 is closely related to autophagy in the process of osteogenic differentiation. During osteoblast differentiation in human osteoblasts (HFOB 1.19), the expression levels of type 2 cannabinoid receptor (CNR2) are increased, which activates autophagy and decreases the expression levels of p62, the latter can compete with NRF2 for binding to KEAP1 (38). CNR2 agonists promote osteogenesis and autophagy, while inhibiting NRF2 entry into the nucleus (39). However, some studies have shown that KEAP1 stabilizes NRF2 via the P62-dependent autophagy degradation pathway (40).

NRF2 has also been shown to promote the differentiation of osteoblasts. The increased ALP activity in osteoblast differentiation and matrix mineralization promotes the levels of free phosphate and stimulates mRNA levels of Nrf2, which may be explained by conferring protection for the cells survival in the harsh environment caused by mineralization (41). Moderate NRF2 activation (Keap1 ^{+ /–} ) or hyperactivation of NRF2 (Keap1^–/– ) enhanced gene expression of Alpl and Wnt5a in primary osteoblasts (42). The expression levels of Runx2 and Col1a1 were decreased in Nrf2 knockout primary osteoblasts (43). Another study showed that the number of osteoblasts colony-forming units and the mineralization of primary calvarial osteoblast from Nrf2^−/− mice were significantly reduced compared with wild type (44). In addition, a study have indicated that the proliferation and differentiation of primary osteoblasts derived from Nrf2^−/− mice are not significantly changed (45).

Certain drugs can alleviate oxidative stress-induced osteogenic differentiation damage by activating NRF2. Aucubin can alleviate the H₂O₂-induced osteogenic differentiation inhibition via the NRF2 signaling pathway (46). Bach1 is a competitive inhibitor of NRF2, which impairs cellular activity and function of MC3T3-E1 cells. BACH1 silencing can inhibit the production of reactive oxygen species (ROS) and promote osteoblast differentiation by enhancing the NRF2/ARE signaling pathway (47). Aucubin, gastrodin, Z-guggulsterone, proanthocyanidins and can prevent the osteogenic differentiation damage induced by glucocorticoids by activating the NRF2 pathway (46, 48–50) ( Figure 3 ).

Many chemicals alleviate oxidative stress-induced cytotoxicity by activating the NRF2 signaling pathway in osteoblasts. For example, luteolin, melatonin and glabridin can ameliorate the methylglyoxal or high glucose-induced cytotoxicity in osteoblasts by activating the NRF2/ARE signaling pathway (51–53). Sulforaphane, indole-3-carbinol, plumbagin and alpinumisoflavone can ameliorate glucocorticoid-induced cytotoxicity of osteoblasts by activating the NRF2 signaling pathway (54–57). Dex significantly reduced the half-life of the cell cycle-related protein P21WAF1/CIP1 by inhibiting the AKT signaling pathway in osteoblasts. Depletion of P21 inhibits NRF2/HO-1 signaling pathway, increasing ROS production and osteoblast damage (58). Chlorogenic acid plays a osteogenic protective role by relieving the inhibition of Dex on P21 and by activating the NRF2 signaling pathway (59). The AKT activator SC79, Icariside II and FGF23 can alleviate the Dex-induced cytotoxicity by activating the AKT-NRF2 signaling pathway (60–62). miR-107 and the AMPK activator compound 991 can activate AMPK-NRF2 signaling pathway to resist Dex-induced oxidative stress damage (63, 64). These results indicate that the AKT-NRF2 and AMPK-NRF2 signaling pathways play an important role in these processes. In addition, inhibition or depletion of phosphoglycate kinase 1 (PGK1) leads to accumulation of methylglyoxal, which modifies KEAP1 to activate the NRF2 pathway and protect against Dex-induced damage in osteoblasts (65). δ-tocotrienol, stilbene glycoside, mangiferin and fermented oyster extracts can relieve H₂O₂-induced damage by activating the NRF2 pathway in osteoblasts (66–69). Exian decoction can reduce the damage of TNF-α to osteoblasts by activating the AKT-NRF2 pathway (70). In addition, A phenolic acid phenethyl urea derivative can ameliorate the radiation damage by activating the NRF2 pathway (71).

The role of NRF2 in osteogenic differentiation of stem cells remains controversial. The colony formation ability of bone marrow mesenchymal stem cells (BMSCs) derived from Nrf2^−/− mice was significantly reduced, resulting in a significant reduction in the number of osteoblasts in 3-week-old Nrf2^−/− mice (45).Nuclear abundance of NRF2 was decreased during adipogenesis of mouse bone marrow stromal cells (ST2 cells), which indirectly demonstrated the promoting effect of NRF2 on osteogenic differentiation of stem cells (72). In addition, hypoxic environment is conducive to the maintenance of stemness and the osteogenic ability of BMSCs. Nrf2 knockout resulted in impaired stemness and osteogenic ability of BMSCs under hypoxia. It has been suggested that NRF2 activation may be a potential method to maintain the stemness of BMSCs under hypoxia conditions (73).

NRF2 may also have a negative effect on osteogenic differentiation of stem cells. Pretreatment of the human periodontal ligament fibroblasts (hPLFs) with the Wnt agonist lithium chloride (LiCl) or Wnt1 can alleviate H₂O₂-induced osteogenic differentiation and mineralization damages with almost complete inhibition of NRF2, and siRNA-mediated NRF2 inhibition restored these injuries (74).

A certain concentration of ROS can promote bone formation and the role of NRF2 in this process is contradictory. Osteogenic differentiation of adipose derived stem cells (ADSCs) was accompanied by NRF2 downregulation and autophagy activation, ROS produced by H₂O₂ promoted osteogenic differentiation of ADSCs, and further up-regulated of autophagy and NRF2 levels. The effects were enhanced by inhibition of NRF2 and reversed by inhibition of autophagy (75). In contrast, mild oxidative stress induced by low concentration of glucose oxidase (GO) promotes the expression levels of NRF2 and HO-1 in embryonic stem cells (ES), while enhancing the osteogenic differentiation and mineralization, which is reversed by inhibition of NRF2 (76). Deferroamine (DFO) promotes osteogenic differentiation of human periodontal ligament cells (hPDLCs) by increasing ROS concentration, accompanied by the activation of NRF2, and the effects were attenuated by NRF2 inhibition (77).

Many drugs or chemicals can promote osteogenic differentiation of stem cells by the activation of the NRF2 signaling pathway. Dendrobium officinale polysaccharides (DOPs) significantly promoted osteoblast differentiation and inhibited adipogenic differentiation of BMSCs by activating NRF2 pathway (78). The novel neohesperidin dihydrochalcone analogue inhibits adipogenic differentiation of human adipose-derived stem cells via the NRF2 pathway as well (79). Panax ginseng fruit protects against porphyromonas gingivalis lipopolysaccharide (PG-LPS)-induced osteoblastic differentiation injury in hPDLCs by activating the NRF2/HO-1 signaling pathway (80). Metformin alleviates H₂O₂-induced osteogenic differentiation damage of periodontal ligament stem cells (PDLSCs) by activating NRF2 signaling pathway (81). The inhibition of nuclear import of NRF2 by Ochratoxin A (OTA) resulted in the loss of the maintenance of self-renewal capacity and differentiation potential in the osteogenic lineage of human mesenchymal stem cells (hMSCs), while enhanced nuclear import of NRF2 by tert-butylhydroquinone (t-BHQ) had opposite effects (82).

The protective effect of NRF2 activation on stem cells has been confirmed by several studies. The interference of KEAP1 expression in adipo-derived mesenchymal stem cells (AD-MSCs) by using the CRISPR/Cas9 method, or directly over expressed NRF2, which can significantly enhance the anti-oxidative ability of stem cells (83, 84). Estradiol (E2) activates NRF2/SIRT/Mn SOD3 pathway in human umbilical cord blood mesenchymal stem cells (hUCB-MSCs), which protects cells from high glucose-induced mtROS production and autophagic cell death (85). On the other hand, Triclosan (TCS) induces cytotoxicity via disrupting the nuclear translocation of SKN-1/NRF2 in hMSCs (86). Polydatin, Coenzyme Q10 and Fufang Lurong Jiangu capsule can effectively prevent the activity decline of BMSCs induced by H₂O₂ via activating the NRF2 pathway (87–89). In addition, BMSCs pretreated with a low dose of non-toxic concentrations of H₂O₂ enhanced the ability of BMSCs to resist oxidative stress injury induced by high concentrations of H₂O₂ by activating the NRF2 signaling pathway, which demonstrated the central role of NRF2 in the electrophile counterattack response (22, 90).

Bone acquisition or loss, the number and activity of osteoblasts and osteoclasts in Nrf2^–/– animal models are controversial. Most studies have shown that Nrf2 knockout impairs animals bone. NRF2 is critical for animal bone development after birth. In female Nrf2^–/– mice, increased oxidative stress significantly inhibits the colony formation ability of BMSCs, resulting in impaired osteoblast formation, manifested as significantly impaired bone acquisition at 3 weeks of age. The bone mass gradually returning at 12 weeks. The contribution of other antioxidant pathways in vivo that are not dependent on NRF2 signaling, such as FOXO, may compensate for the loss of NRF2 (45). As previously mentioned, several studies have shown that NRF2 regulation of RANKL/OPG does not cause significant difference in mouse osteoclast numbers and activity (29, 30, 45). However, significant upregulation of osteoclast numbers and activity in Nrf2^–/– mice has also been observed in several studies (95, 124). Female Nrf2^–/– mice aged 6 to 9 months exhibited higher osteoblast numbers and activity than WT mice, however, the increased RANKL/OPG ratio resulting in the more osteoclast numbers, the stronger activity of osteoclasts and the bone conversion rate, which ultimately led to bone loss (95). Compared with age-matched WT mice, bone formation and mineralization in male Nrf2^–/– mice at 17 weeks of age were significantly decreased, and the numbers and surface of osteoclasts were significantly increased, leading to severe bone loss (124).

NRF2 inhibition has also been observed to benefit bone acquisition in animals. Nrf2^–/– mice at 9 weeks of age exhibited higher bone mass, higher mineral attachment rate, higher level of serum osteocalcin and the number of osteoblasts than WT mice (30).

In addition, a study showed that the number of osteoblasts decreased and the number of osteoclasts increased in sex-matched Nrf2^–/– mice at 3 months of age, but bone volume did not change significantly (44). Another study also found that bone parameters of Nrf2^–/– female mice were similar to those of WT mice (94).

NRF2 may be indispensable in the protection of bones against environmental stress damage. Irradiation (IR) induced more severe bone loss in Nrf2^–/– mice compared to WT mice. Further studies have shown that NRF2 plays a critical role in osteoblast differentiation and stromal formation regardless of IR presence (44). NRF2 inhibition reduce load-driven bone formation. The parameters, such as femur bone mineral density, bone formation rate and ultimate force as well as relative mineralizing surface and relative bone formation rate were significantly reduced in Nrf2^–/– mice (124). NRF2 plays an important role in bone regeneration. NRF2 is activated in the process of fracture healing. Nrf2^–/– mice suffered from more brittle bone cortex and significantly reduced anti-fracture force and anti-bending stress. Nrf2-knockout induced aggravation of oxidative damage of cartilaginous callus in femoral shaft fracture model mice, resulting in delayed bone healing and remodeling. It is worth mentioning that the expression of osteocalcin mRNA in the damaged tissues of Nrf2^–/– mice was significantly lower than that of the WT mice, indicating that NRF2 binding to the osteocalcin promoter caused upregulation of osteocalcin expression in vivo (125), which is contradicted in previous study (27).

In addition, a study explored the differences in the effects of Nrf2-knockout on bone in male and female, old and young mice. The regulation of NRF2 on bone mass accumulation depends on sex and it is necessary in females, while dispensable or even deleterious in males. The total, femoral and spinal bone mineral density (BMD) of Nrf2^–/– female mice was lower than that of female WT mice for at least a period of time, however, NRF2 deficiency inhibited bone loss in old male mice, suggesting that it negatively affected bone maintenance in male bones. In both young and old female mice, complete endogenous antioxidant responses (phase II detoxifying and antioxidant enzyme expression) were undertaken by NRF2-dependent mechanisms, whereas in older male mice, It was entirely undertaken by NRF2-independent mechanisms. And the NRF2-dependent phase II detoxifying enzyme expression and NRF2-independent oxidase expression coexist in male young mice. It means that the maintenance of the optimal antioxidant response in female bones depends on NRF2, whereas the aging skeleton of males does not require NRF2 for protection against oxidative stress (126).

Activation of NRF2 is critical for osteocytogenesis, through direct activation of certain osteocyte-specific genes (Dmp1, Mepe, and Sost) and osteoblastic gene (Runx2 or Osx). Both sexes of osteocytic cell lineage-specific Nrf2^–/– mice (Dmp1/Nrf2-KO) showed severe osteopenia. However, Osteoblast-specific Nrf2^–/– mice (Col1a1/Nrf2-KO) indicated subtle osteopenia only in males, whereas no significant changes were noted in female bones. Col1a1/Nrf2-KO males and females showed the reduced expression of osteocyte-specific genes (Dmp1, Mepe, and Sost), but no significant changes in expression of osteoblastic genes (Runx2 or Osx). The differences in osteogenesis gene expression may be due to the deletion period of Nrf2 in osteocytic cell lineage is later than that in osteoblasts (Late embryo and neonatal period), the Col1a1/Nrf2-KO mice may have some compensatory mechanisms to restore the bone phenotypes. The number of osteoclasts was significantly increased in males but not females in both Dmp1/Nrf2-KO and Col1a1/Nrf2-KO mice. The sex difference complicated the effect of Nrf2-knockout in osteogenic lineages on paracrine secretion of osteoclasts by regulating RANKL/OPG ratio (43).

NRF2 hyperactivation neonatal mice constructed from homozygous Keap1 mutation died within three weeks due to severe hyperkeratosis in the esophagus and forestomach (127). One study observed that male neonatal Keap1^–/– mice showed slight talus and calcaneus deformation and shortened growth plate hypertrophy, and the number of osteoclasts was decreased (29). Viable Keap1^–/– mice were generated by the deletion of Nrf2 in esophageal (NEKO mice), the data indicated that NEKO mice at the age of 8-10 weeks exhibited small body size and low bone density. As the differentiation of both osteoclasts and osteoblasts was weakened, it was suggested that the impaired differentiation of osteoblasts but not osteoclasts was the cause of bone hypoplasia in NEKO mice (28).

Moderate activation of NRF2 by heterozygous Keap1 mutation(Keap1^+/− ) produced another viable Nrf2-overexpressed mice model. The bone phenotype of NRF2 moderate activation indicated sexual dimorphism. Bone formation was significantly increased and bone resorption was significantly reduced in male Keap1^+/− mice compared to the littermate controls. And no significant effects were noted in females. In addition, The male Keap1^+/− and Keap1^−/− neonatal mice showed more bone mineralization in their limbs than WT mice, suggesting that hyperactivation and moderate activation of NRF2 could both accelerate bone mineralization (42). However, osteogenic differentiation was impaired in NEKO mice (28). In the future, osteoblast-specific Keap1^−/− mice may aid to further elucidate differences in the role of KEAP1/NRF2 in bone formation (42). Briefly, more studies show that hyperactivation of NRF2 may cause damage to bone formation, while moderate activation of NRF2 promotes increased bone mass.