CHIR-99021 regulates mitochondrial remodelling via β-catenin signalling and miRNA expression during endodermal differentiation

Mitochondria are highly specialized and dynamic organelles inside mammalian cells that play crucial roles in several cellular pathways, such as the production of energy in the form of ATP via oxidative phosphorylation, apoptosis and reactive oxygen species (ROS) signalling (Friedman and Nunnari, 2014; Chandel, 2015). To meet the different energy demands of distinct cell types, cells modulate mitochondrial numbers and activity through biogenesis and through dynamic events (Eisner et al., 2018). Recently, mitochondria have been closely linked to cell fate determination and development, and evidence has demonstrated that mitochondrial biogenesis is essential for the successful differentiation of stem cells (Varum et al., 2011; Chen et al., 2012; Folmes et al., 2012; Michel et al., 2012; Xu et al., 2013; Wanet et al., 2015). An increased mitochondrial mass and mitochondrial DNA (mtDNA) copy number, together with the elongation of the mitochondrial network and maturation of the cristae ultrastructure, has been observed during the differentiation of stem cells in vitro (Chen et al., 2008; Varum et al., 2011; Wanet et al., 2014, 2017). The morphology, localization, abundance and function of mitochondria could potentially be used as key markers for the differentiation of stem cells into specialized cell types (Varum et al., 2011; Wanet et al., 2017). However, little is known about the regulatory mechanisms connecting mitochondrial structural and functional remodelling to the differentiation of stem cells. Thus, a better understanding and control of mitochondrial remodelling during the differentiation process in the laboratory should translate into enhanced efficiency and increased fidelity in the resulting cells.

Previously, we reported that the pharmacological inhibition of glycogen synthase kinase-3α and -3β (GSK-3α, GSK-3β, also known as GSK3A and GSK3B) with specific inhibitors (CHIR-99021 and CHIR-98014) initiates the efficient differentiation of human adipose stem cells (hASCs) to human definitive endodermal progenitor cells (hEPCs) by upregulating the transcription factors GATA4, FOXA2 and SOX17 via activation of the Wnt/β-catenin pathway (Huang et al., 2017). CHIR-99021 has also been reported as an important cocktail component of endodermal differentiation inducers for pluripotent stem cells (Blauwkamp et al., 2012; Lian et al., 2014; Teo et al., 2014; Bao et al., 2015; Morrison et al., 2016). CHIR-99021-induced GSK-3 inhibition during differentiation further resulted in the inhibition of T cell factor 3 (TCF3, also known as transcription factor 7-like 1, Tcf7l1), which in turn relieved the repression of FoxA2 in mouse embryonic stem cells (Morrison et al., 2016). The inhibition of TCF4 (also known as Tcf7l2), another member of the transcription factor 7-like family, was sufficient to increase the expression of peroxisome proliferator-activated receptor gamma (PPARγ) coactivator-1α (PGC-1α, also known as PPARGC1A), the master regulator of mitochondrial biogenesis (Ventura-Clapier et al., 2008) and oxidative phosphorylation activity during the early step of human bone marrow mesenchymal stem cell commitment to hepatic differentiation (Wanet et al., 2017). We therefore anticipated that the effects of CHIR-99021 on endodermal induction may be attributable to the recapitulation of mitochondrial biogenesis and glycolysis-oxidative phosphorylation shifts.

Mitochondrial biogenesis and oxidative phosphorylation activity are accomplished by the coordinated expression of genes from the nuclear and mitochondrial genomes. Several transcription and/or replication factors have been reported to directly regulate gene expression in mitochondrial biogenesis. Nuclear respiratory factor 1 (NRF1, also known as NFE2L1) and NRF2 (also known as NFE2L2) control all ten nucleus-encoded cytochrome oxidase subunits, and the orphan nuclear hormone receptor, ERRα (also known as ESRRA), controls the medium chain acyl-coenzyme A dehydrogenase (MCAD) promoter (Huss and Kelly, 2004). The PGC-1 family of regulated coactivators [PGC-1α, PGC-1β (also known as PPARGC1B) and PRC (also known as PPRC1)] targets multiple transcription factors, including NRF1, NRF2, and ERRα plays a central role in a regulatory network governing the transcriptional control of mitochondrial biogenesis (Wu et al., 1999; Scarpulla, 2011; Wanet et al., 2017). PGC-1 binds to and coactivates the transcriptional function of NRF1 on the promoter for mitochondrial transcription factor A (TFAM), a direct regulator of mitochondrial DNA replication and/or transcription (Virbasius and Scarpulla, 1994; Picca and Lezza, 2015). Hence, CHIR-99021 may regulate mitochondrial biogenesis and activities, at least in part, through modulating the amounts and/or functions of these key sets of regulators. Here, we investigated whether CHIR-99021 can regulate certain mitochondrial regulatory factors to control mitochondrion biogenesis and oxidative phosphorylation activities during the definitive endodermal differentiation of hASCs.

Previously, we reported that the efficient differentiation of hASCs to hEPCs was initiated by activin A, Wnt3a, or the pharmacological inhibition of GSK-3 (CHIR-99021 and CHIR-98014) separately. The hEPCs were directed to differentiate synchronously into hepatocyte-like cells after further combinations of soluble factors by a reproducible three-stage method (Li et al., 2014; Huang et al., 2017). Among these factors, CHIR-99021 exhibited high efficiency for upregulating endodermal-specific transcription factors, including FOXA2, GATA4, SOX17, and C-X-C motif chemokine receptor 4 (CXCR4) (Huang et al., 2017).

Definitive endoderm is the common progenitor to epithelial cells in diverse internal organs, including the liver, pancreas and intestines (Tremblay and Zaret, 2005). To examine their multilineage differentiation potential, we cultured hEPCs in pancreatic (Aigha et al., 2018) or intestinal (Spence et al., 2011) induction medium for four days and then evaluated the cells for expression of pancreatic lineage-specific transcription factors pancreatic and duodenal homeobox1 (PDX1) (Offield et al., 1996) and caudal type homeobox 2 (CDX2), a member of the caudal-related homeobox transcription factor gene family. CDX2 is a major regulator of intestine-specific genes involved in cell growth and differentiation (Gao et al., 2009). Immunocytochemistry analyses confirmed that expression levels of PDX1 in pancreatic lineage differentiated cells (Fig. S1A) and of CDX2 in intestinal lineage differentiated cells (Fig. S1B) were significantly increased compared to hASCs. These data confirmed that hEPCs induced by treatment with CHIR-99021 possess multilineage differentiation potential.

To investigate mitochondrial biogenesis during definitive endodermal differentiation of hASCs, the expression of the transcription factors PGC-1α, TFAM and NRF1 in hEPCs and hASCs was determined. The results showed that the mRNA levels of PGC-1α, TFAM and NRF1 in hEPCs were significantly increased compared to hASCs (Fig. S2). Next, two fluorescent probes were used to assess the mitochondrial mass (MitoTracker Green) and mitochondrial membrane potential (TMRM) in hASCs and hEPCs. The relative intensities of MitoTracker Green FM and TMRM were calculated, and the relative intensity of TMRM fluorescence was normalized to MitoTracker Green FM in each cell. The results showed that the mitochondrial mass and mitochondrial membrane potential in the hEPCs were significantly higher than those in the hASCs (Fig. 1A,B).

To further demonstrate changes in mitochondrial numbers and structures, we performed ultrastructural analyses using transmission electron microscopy (TEM). Upon definitive endodermal differentiation, the hEPCs showed a higher number and density of mitochondria than did the hASCs (Fig. 1C,D). Mitochondria in hASCs contained functionally immature mitochondria with a globular shape, poorly developed cristae and perinuclear localization, all indicative of a less-active mitochondrial state, while the hEPCs possessed a complex morphology, with developed cristae, a denser matrix, and an elongated appearance (Fig. 1E,F).

To investigate aerobic metabolism activity levels during definitive endodermal differentiation, we quantified ATP production rates by measuring the two major production pathways in mammalian cells; glycolysis and oxidative phosphorylation were analysed using Agilent Seahorse XF technology. Differentiating cells undergo a metabolic switch from highly glycolytic metabolism to active mitochondrial aerobic metabolism and drive ATP production though oxidative phosphorylation (Khacho and Slack, 2017). As shown in Fig. 1G, the ATP production rate in hEPCs was higher than that in hASCs. Moreover, the proportional mitochondrial oxidative phosphorylation pathway ATP (mitoATP) production rate in hEPCs was ∼63.4%, while the proportional mitoATP production rate in hASCs was only ∼29.2%. The proportional glycolytic pathway ATP (glycoATP) production rate in hEPCs was ∼36.6%, while the proportional glycoATP production rate in hASCs was ∼71.8%. The increased oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) values in hEPCs represented mitochondrial oxidative phosphorylation instead of glycolysis (Fig. 1H,I). This finding indicates that ATP production in hEPCs mainly depends on oxidative phosphorylation in mitochondria, while ATP production in hASCs mainly depends on glycolysis. These results suggest that mitochondrial biogenesis and oxidative phosphorylation capacities are increased as cells undergo a cell fate transition during the definitive endodermal differentiation of hASCs.

Growing evidence suggests that the mitochondrial dynamics involved in fission and fusion enable mitochondria to divide and help ensure proper organization of the mitochondrial network during biogenesis (Ventura-Clapier et al., 2008). To investigate the role of mitochondrial dynamics in definitive endodermal differentiation of hASCs, mitochondrial fission was impaired with genetic inhibition of dynamin 1­-like (DRP1, also known as DNM1L) (Agarwal et al., 2016) in hASCs, prior to the onset of endodermal differentiation. As shown in Fig. S3, the mRNA and protein level of DRP1 in cells treated with siRNA targeting DRP1 were successfully reduced by 83.1% and 74.9%, respectively, compared to the control siRNA group. After reduction of the DRP1 expression level, morphology of the mitochondria became long and thin (Fig. 2A) during definitive endodermal differentiation. The mitochondrial mass and membrane potential in cells treated with DRP1 siRNA were significantly lower than those in the control siRNA cells (Fig. 2B). Meanwhile, inhibition of DRP1 significantly decreased the mRNA levels of PGC-1α, TFAM and NRF1 (Fig. 2C), and protein levels of TFAM in DRP1 siRNA-treated cells were decreased compared to levels in control siRNA-treated cells (Fig. 2D). This result indicates that inhibition of DRP1 significantly affects mitochondrial dynamics and mitochondrial biogenesis during definitive endodermal differentiation.

Furthermore, mRNA levels of FOXA2, GATA4 and SOX17 in DRP1 siRNA-treated cells were significantly decreased compared to those in control siRNA-treated cells (Fig. 2E). Immunocytochemistry analyses showed that the relative fluorescence intensities of GATA4 in DRP1 siRNA-treated cells were also significantly decreased compared to control siRNA-treated cells (Fig. 2F). These results suggest that mitochondrial dynamics play a key role in the definitive endodermal differentiation of hASCs.

To investigate the role of CHIR-99021 in mitochondrial biogenesis and oxidative phosphorylation during differentiation, hASCs were treated with 2 µM CHIR-99021 or vehicle control (DMSO) for 24 h. Insulin-transferrin-selenium (ITS) was then added to the medium for another 48 h. The properties of mitochondrial biogenesis were first compared between the two groups. The results showed that the mitochondrial mass and mitochondrial membrane potential were significantly increased in CHIR-99021-treated cells compared to vehicle control-treated cells (Fig. 3A,B). The mitochondrial morphology of CHIR-99021-treated cells exhibited developed cristae and a more dense matrix compared to the vehicle control-treated cells (Fig. 3C,D).

To further demonstrate changes to the mitochondrial components of DNA and protein, mtDNA copy numbers and protein subunits of the respiratory chain were assessed in CHIR-99021-treated cells and vehicle control-treated cells. The mtDNA content in CHIR-99021-treated cells increased ∼1.57-fold over the content in vehicle control-treated cells (Fig. 3E). The mRNA levels of cytochrome c oxidase (COX) subunit 6C (COX6C) and COX8A in CHIR-99021-treated cells were significantly higher than in the vehicle control-treated cells (Fig. 3F). The protein levels of COX subunit 1 (also known as MTCO-1) (Dennerlein and Rehling, 2015), which is mtDNA-encoded, and subunit A of mitochondrion complex II (succinate dehydrogenase complex flavoprotein subunit A, SDHA) (Bezawork-Geleta et al., 2018), which is nDNA-encoded, were both significantly increased in CHIR-99021-treated cells compared to levels in vehicle control-treated cells (Fig. 3G).

As shown in Fig. 3H, the real-time ATP production rate in CHIR-99021-treated cells was higher than that in the vehicle control-treated cells. However, there was no difference in the proportional mitoATP production rate between the two groups. The proportional mitoATP production rate in CHIR-99021-treated cells was ∼63.4%, while the proportional mitoATP production rate in the vehicle control-treated cells was ∼59.37%. The OCR and ECAR values in CHIR-99021-treated cells were also higher than those in the vehicle control-treated cells (Fig. 3I,J).

The production of ROS in CHIR-99021-treated cells was significantly higher than in the vehicle control-treated cells after tert-butyl hydrogen peroxide (TBHP) treatment (Fig. S4A). These data show that CHIR-99021 upregulates mitochondrial biogenesis and oxidative phosphorylation activities during definitive endodermal differentiation of hASCs.

In an attempt to understand the mechanism by which CHIR-99021 promotes mitochondrial biogenesis and oxidative phosphorylation activities, we next examined the expression of transcription factors PGC-1α, TFAM and NRF1 in CHIR-99021-treated hASCs and vehicle control-treated hASCs. These transcription factors are involved in mitochondrial biogenesis and oxidative phosphorylation activities. Real-time RT-PCR analyses showed that mRNA levels of PGC-1α, TFAM and NRF1 in CHIR-99021-treated cells were significantly higher than those in the vehicle control group (Fig. 4A). Western blotting analyses showed that the protein level of PGC-1α in CHIR-99021-treated cells was higher than that in the vehicle control-treated group (Fig. 4B). Immunocytochemistry analyses showed that the relative fluorescence intensities per cell of TFAM (Fig. 4C) and NRF1 (Fig. 4D) in CHIR-99021-treated cells were also higher than those in the vehicle control-treated cells. These findings suggest that CHIR-99021 increases the expression of transcription factors involved in mitochondrial biogenesis and oxidative phosphorylation activities.

To assess whether the effect of CHIR-99021 on the induction of mitochondrial remodelling depends on the β-catenin signalling pathway, we disrupted the signalling pathway by delivering siRNA to knock down expression of the β-catenin gene in hASCs. As shown in Fig. S5, the mRNA and protein levels of β-catenin in β-catenin siRNA-treated cells were successfully reduced by 83.1% and 50.1%, respectively, compared to the control siRNA-treated group. After reducing β-catenin expression using siRNA, mitochondrial biogenesis and oxidative phosphorylation activities were assessed in β-catenin siRNA-treated and control siRNA-treated cells. The results showed that mitochondrial mass and mitochondrial membrane potential were significantly decreased in β-catenin siRNA-treated cells compared to control siRNA-treated cells (Fig. 5A,B). The mitochondrial morphology in β-catenin siRNA-treated cells exhibited a loss of integrity of the outer membrane and undeveloped cristae compared to the control siRNA-treated cells (Fig. 5C,D). The mtDNA content in β-catenin siRNA-treated cells showed a decrease of ∼25% compared to the control siRNA-treated cells (Fig. 5E). The protein levels of SDHA and MTCO-1 significantly decreased in the β-catenin siRNA-treated cells compared to those in the control siRNA-treated cells (Fig. 5F).

As shown in Fig. 5G, the ATP production rate in the β-catenin siRNA-treated cells was slightly lower than in control siRNA-treated cells. The proportional mitoATP production rate in β-catenin siRNA-treated cells was ∼65.2%, while the proportional mitoATP production rate in vehicle control-treated cells was ∼72.9%. The OCR values in the β-catenin siRNA-treated cells were also lower than those in the vehicle control-treated cells (Fig. 5H,I). However, the production of ROS in β-catenin siRNA-treated cells was not different to control siRNA-treated cells after treatment with TBHP (Fig. S4B). These findings suggest that the CHIR-99021 regulation of mitochondrial biogenesis and oxidative phosphorylation activities during definitive endodermal differentiation of hASCs depends on the β-catenin signalling pathway.

To further assess whether the induction of transcription factors by CHIR-99021 depends on the β-catenin signalling pathway, mRNA and protein levels of PGC-1α, TFAM and NRF1 were determined in β-catenin siRNA-treated and control siRNA-treated hASCs. The results showed that the mRNA and protein levels of NRF1 in β-catenin siRNA-treated cells were decreased compared to those in the control siRNA-treated cells (Fig. 6A,B). The mRNA level of PGC-1α was not changed, but the protein level of PGC-1α in β-catenin siRNA-treated cells was decreased compared to control siRNA-treated cells (Fig. 6A,C). While the mRNA level of TFAM was increased in β-catenin siRNA-treated cells, the protein level was decreased compared to control siRNA-treated cells (Fig. 6A,D). These findings indicate that the regulation of transcription factors by CHIR-99021 through increasing the expression of mRNAs during the definitive endodermal differentiation of hASCs partly depends on the β-catenin signalling pathway.

MicroRNAs (miRNAs) are key components of a broadly conserved post-transcriptional regulatory mechanism that controls gene expression by targeting mRNAs (Ong et al., 2015). Previous research has found that CHIR-99021 decreased the level of mature miRNAs of most miRNA species in mouse embryonic stem cells (ESCs) (Wu et al., 2015). To further decipher the possible role of CHIR-99021 in miRNA expression, we investigated the effects of miRNAs on the regulation of transcription factors. Six candidate miRNAs that potentially target the 3′-UTR of PGC-1α were predicted by TargetScan and PicTar (Table S3), and were selected for further investigation using qRT-PCR assay. The results showed that the expression levels of miR-19b-2-5p, miR-23a-3p, miR-23c, miR-130a-3p and miR-130a-5p (Fig. S6A) were significantly decreased in CHIR-99021-treated cells. Among these, miR-19b-2-5p was also found to target key mitochondrial biogenesis transcription factors, including TFAM and NRF1 (Table S4, Fig. S6B) (Saini et al., 2018). These results suggest that miR-19b-2-5p may be a key miRNA that regulates mitochondrial biogenesis. These results also suggest that CHIR-99021 regulates mitochondrial biogenesis and oxidative phosphorylation activities by increasing the expression of transcription factors, a process that is mediated by miRNAs and in a β-catenin-dependent manner. Understanding the mitochondrial biogenesis and metabolic regulation of stem cell differentiation will have great value in manipulating the increased efficiency of cellular differentiation potential.

The crucial role of mitochondria and the bioenergetic function in stem cells has recently come to light, and the importance of mitochondria in mediating stem cell fate is increasingly recognized (Wanet et al., 2015; Zhang et al., 2018). The precise mechanism by which mitochondrial remodelling is regulated during stem cell differentiation has not been elucidated. In the present study, we found that CHIR-99021 promotes mitochondrial biogenesis and oxidative phosphorylation activities during definitive endodermal differentiation of hASCs via activation of the β-catenin signalling pathway. We also demonstrated that CHIR-99021 decreases the expression of miR-19b-2-5p, miR-23a-3p, miR-23c, miR-130a-3p, and miR-130a-5p in hEPCs differentiated from hASCs. These results suggest that CHIR-99021 may be one of the key factors that mediates mitochondrial structural and functional remodelling during the differentiation of stem cells in vitro.

As sites of cellular respiration and energy production, mitochondria display a characteristic ultrastructure. We observed elongation of the mitochondrial network and maturation of the cristae ultrastructure in hEPCs differentiated from hASCs. Healthy mitochondrial membranes maintain a difference in electrical potential between the interior and exterior of the organelle, referred to as membrane potential. TMRM analysis showed that the mitochondrial membrane potentials in hEPCs were higher than those in hASCs.

In addition to morphological and ultrastructural changes, we found that hEPCs exhibited elevated active mitochondrial metabolic function. Mitochondria are notably recognized for their generation of ATP through oxidative phosphorylation. In addition to ATP generation through mitochondrial respiration, the formation of ATP through glycolysis is an essential cellular energy metabolic process in cells. Energy production in stem cells mainly results from a high glycolysis rate, leading to increased lactate production, whereas oxidative phosphorylation activity is limited, leading to reduced oxygen consumption and ROS production (Wanet et al., 2015). Therefore, measuring the rate of ATP production from both pathways simultaneously in live cells enables a view into cellular mitochondrial metabolic function that is not provided by simply measuring the amount of ATP level in the cell. We found that hEPCs exhibited elevated oxygen consumption, ATP production predominantly through oxidative phosphorylation activity, and elevated ROS production. In contrast, hASCs exhibited lower oxygen consumption, ATP production predominantly through glycolysis activity, and lower ROS production. These results indicate mitochondrial structural and metabolism remodelling upon definitive endodermal differentiation (Wanet et al., 2015), which is in agreement with data showing that mitochondrial biogenesis occurs during osteogenic differentiation (Chen et al., 2008; Quinn et al., 2013), adipogenic differentiation (Hofmann et al., 2012; Quinn et al., 2013; Zhang et al., 2013) and hepatogenic differentiation of mesenchymal stem cells (Wanet et al., 2014).

Growing evidence suggests that the mitochondrial dynamics involved in fission and fusion enable mitochondria to divide and help ensure proper organization of the mitochondrial network during biogenesis (Ventura-Clapier et al., 2008). Fission is mediated by DRP1 and fission protein 1 (FIS1) (Detmer and Chan, 2007). DRP1, a master fission mediator, is a cytosolic dynamin GTPase that is translocalized to the mitochondrial outer membrane, forming constricting spirals around mitochondria to facilitate mitochondrial division. During differentiation, treatment of hASCs with DRP1 siRNA caused extensive elongation of mitochondria, which coincided with decreased expression of mitochondrial biogenesis transcription factors, including PGC-1α, TFAM and NRF1, and endodermal-specific transcription factors including FOXA2, GATA4 and SOX17. These results suggest an essential role of mitochondrial dynamics in successful definitive endodermal differentiation progression, similar to previous work from Kim et al. (2013) reporting that inhibition of DRP1-dependent mitochondrial division impairs myogenic differentiation.

Based on our results, we propose that CHIR-99021 may participate in the regulation of mitochondrial biogenesis. Next, the properties of mitochondrial biogenesis were compared between CHIR-99021-treated cells and vehicle control-treated cells. The results confirmed that mitochondrial components, including mtDNA copy number, amount of protein subunits and mitochondrial oxidative phosphorylation activity were significantly increased following treatment with CHIR-99021. Mitochondrial biogenesis in eukaryotic cells requires a set of transcription factors (Scarpulla, 2011). Among these, PGC-1α is considered to be a crucial regulator of mitochondrial biogenesis, as it can physically dock with and coactivate transcription factors such as TFAM and NRF1, which modulate the expression of genes encoding mitochondrial proteins (Virbasius and Scarpulla, 1994; Wu et al., 1999; Ventura-Clapier et al., 2008). Consistent with the mitochondrial component and function increase, the expression levels of PGC-1α, TFAM, and NRF1 were increased in CHIR-99021-treated cells. This finding indicates that the increase in mitochondrial biogenesis in hEPCs correlates with upregulation of the expression of key master transcription factors.

Earlier studies showed that lithium increases PGC-1α expression and mitochondrial biogenesis in primary bovine aortic endothelial cells (Struewing et al., 2007) and megakaryocytes (Undi et al., 2017) by regulating β-catenin signalling. To understand the mechanism of CHIR-99021 on the induction of mitochondrial biogenesis, we disrupted the β-catenin signalling pathway by delivering siRNA to knock down the expression of β-catenin in hASCs. The results showed that downregulating the expression of β-catenin impairs the effect of CHIR-99021 on the induction of mitochondrial metabolism and remodelling upon definitive endodermal differentiation of hASCs. However, mRNA levels of PGC-1α and TFAM did not decrease in β-catenin siRNA-treated cells. These findings indicate that there are other factors involved in regulating PGC-1α and TFAM mRNA levels during definitive endodermal differentiation from hASCs, apart from the β-catenin signalling pathway.

During embryonic development, cells within embryos are exposed to numerous ligands and morphogens of several signalling pathways. miRNAs may sharpen morphogen gradients in the developing embryo or serve as a positive regulator by amplifying signal strength in duration to allow cell responsiveness to subthreshold stimuli (Inui et al., 2010). Recent studies have shown that many miRNAs that are involved in the mitochondrial biogenesis of different cell types are thought to regulate the expression of PGC-1α (Ong et al., 2015; Wu et al., 2015; Ai et al., 2016; Cha et al., 2017; Jiang et al., 2017; Portius et al., 2017; Saini et al., 2018). A study found that CHIR-99021 decreased the mature miRNA levels of most miRNA species in mouse ESCs (Wu et al., 2015). In the present study, we found that the expression levels of miR-19b-2-5p, miR-23a-3p, miR-23c, miR-130a-3p, and miR-130a-5p were decreased upon treatment with CHIR-99021. Among these, miR-19b-2-5p targets PGC-1α, TFAM and NRF1, which is consistent with a detailed bioinformatics analysis from Saini et al. (2018). Based on our data and the bioinformatics analysis, we suggest that miR-19b-2-5p could be a key to the regulation of mitochondrial gene expression and biogenesis.

In conclusion, we observed fragmentation of the mitochondrial network morphology, elevated oxidative phosphorylation capacities, and increased production of reactive oxygen species following the pharmacological inhibition of GSK-3α and GSK-3β with CHIR-99021, which initiated the definitive endodermal differentiation of hASCs. We demonstrated that CHIR-99021 promotes mitochondrial biogenesis via activation of mitochondrial biogenesis transcription factors, including PGC-1α, TFAM and NRF1, through the β-catenin pathway, and inhibits the expression of miRNAs during definitive endodermal differentiation of hASCs. Impairment of the mitochondrial dynamic balance or downregulation of β-catenin expression weakens the effect of CHIR-99021 on the induction of mitochondrial remodelling and the expression of key transcription factors for mitochondrial biogenesis. Further research is still required to unravel the mechanisms involved in the interplay between these two pathways. Understanding how these pathways are regulated and determining whether they regulate each other is of great interest to researchers working to improve the directed differentiation of stem cells into specific cell types. Understanding the role played by mitochondria in stem cell differentiation may therefore be useful in regenerative medicine and pharmacological testing.

All human tissues and cells were obtained with donor consent, and experimental protocols were approved and carried out in accordance with the relevant guidelines and regulations of the Ethics Committee of Capital Medical University (2011SY08), China. All clinical investigations were conducted according to the principles expressed in the Declaration of Helsinki.

hASCs from four different donors were developed, as described in our previous report (Li et al., 2014). Cells from passages six to nine were used in this study. Definitive endodermal differentiation was performed as described in our previous report (Huang et al., 2017). Briefly, hASCs were plated on dishes (Nunc) coated with collagen I (Invitrogen) and cultured in DMEM/F-12 (Invitrogen) supplemented with mesenchyme stem cell-screened 10% foetal bovine serum (FBS-MSCS, HyClone), 100 U/ml penicillin and 100 µg/ml streptomycin at 37°C with 5% CO₂. Once the cells reached 90% confluence, they were washed twice with PBS and incubated with serum-free DMEM/F-12 medium for 48 h. The cells were then incubated with DMEM/F-12 containing 0.5 mg/ml albumin fraction V (Sigma-Aldrich) with or without CHIR-99021 (2 µM, Selleckchem) for 24 h, One percent insulin-transferrin-selenium (ITS) (Sigma-Aldrich) was added to the medium beginning on the second day, and cells cultured for 48 h.

For differentiation of hEPCs into pancreatic lineage cells, the cells were incubated with MCDB 131 basal media (Invitrogen) containing 50 ng/ml Noggin (Sigma-Aldrich), 50 ng/ml FGF10 (Peprotech) and 2 µM CHIR-99021 for 2 days. Then the cells were cultured with DMEM medium (Invitrogen) containing 50 ng/ml FGF10, 50 ng/ml Noggin and 1% B27 supplement (Invitrogen) for 2 days.

For differentiation of hEPCs into intestinal lineage cells, the cells were incubated with RPMI 1640 media (Invitrogen) containing 0.5 mg/ml BSA, 1% ITS, 100 ng/ml Activin A (Peprotech), 500 ng/ml FGF4 (Peprotech) and 3 µM CHIR-99021 for 4 days.

hASCs were plated at 20,000 cells/cm² in antibiotic-free basal medium 24 h prior to transfection. siRNA transfection was performed following the manufacturer's protocol, as previously described (Huang et al., 2017). Briefly, ON-TARGET SMARTpool siRNAs directed against β-catenin (L-003482-00-0005, Dharmacon) or non-targeting siRNAs (D-001810-10-05, Dharmacon) were mixed with Transfection DharmaFECT 1 (Dharmacon). DRP1 siRNA (5′-GGCCAAUAGAAAUGGAACATT/UGUUCCAUUUCUAUUGGCCTT-3′) or control siRNA (5′-UUCUCCGAACGUGUCACGUTT/ACGUGACACGUUCGGAGAATT-3′) (SyngenTech) were mixed with Lipofectamine RNAiMAX Transfection Reagent (Invitrogen). After a 20-min incubation at room temperature, the complexes were added to the cells at a final siRNA concentration of β-catenin (25 nM), DRP1 (50 nM). The medium was replenished with antibiotic-free basal medium 24 h post-transfection. The culture medium was changed every 2 days for the duration of the experiment.

Cells were seeded into confocal chambers (NEST) at a density of 40,000 cells per well. For mitochondria and nuclei staining, the cells were incubated with Hoechst 33342 (0.2 µg/ml, Thermo Fisher Scientific) for 20 min at 37°C, then 25 nm TMRM (I34361, Thermo Fisher Scientific) or 50 nM MitoTracker Green FM (M7514, Thermo Fisher Scientific) (Karbowski and Youle, 2003; Medeiros, 2008) was added. The cells were incubated for 30 min at 37°C and then washed three times with PBS. After washing, the images of fluorescently labelled mitochondria and nuclei were captured using a confocal laser scanning microscope (Leica). The emission wavelengths for the individual dyes were 490 nm for MitoTracker Green FM and 574 nm for TMRM. The density of mitochondria in each cell was assessed using ImageJ as described previously (Arena et al., 2017). The data are presented as mean±s.d.

The relative intensities of MitoTracker Green FM and TMRM were calculated in each cell using ImageJ software. ImageJ was used to set mitochondria as the first threshold and remove the influence of cytosolic or background signals. Pictures were converted into bitmap as a binary colour (colour=1, no colour=0). The value of the calibration or scale was adjusted. The threshold was adjusted to separate the red in the picture from the background. Different ranges were selected depending on the outline of every cell. At least 200 cells per group of each parameter were compared among the treatment groups by unpaired two-tailed Student's t-tests using the statistical software SPSS 11.5 (IBM Corporation). The calculation method was performed as previously described (Shprung et al., 2009).

Immunofluorescence analysis was performed as previously described (Li et al., 2014). Briefly, the cells were fixed with 4% paraformaldehyde for 20 min at room temperature, followed by permeabilization with 0.3% Triton X-100 in PBS for 5 min. The cells were rinsed and blocked with 10% goat serum (Zsgb-Bio) for 60 min at room temperature. The cells were then incubated with the primary antibodies, which are listed in Table S2, at 4°C overnight. Following three 5-min washes in PBS with gentle agitation, an Alexa Fluor-conjugated secondary antibody (Invitrogen) at 1:500 was added, and the samples were incubated for 1 h at 37°C. The nuclei were counter-stained with 4′, 6-diamidino-2-phenylindole (DAPI, Sigma-Aldrich).

Quantitative real-time RT-PCR was performed as previously described (Guo et al., 2017). Total cellular RNA was extracted from 2.0×10⁵ cells with the RNeasy Mini Kit (QIAGEN), according to the manufacturer's instructions. For PCR analysis, 0.5 μg of RNA was reverse-transcribed to cDNA using Superscript III reverse transcriptase and random hexamer primers (Invitrogen). Real-time PCR analysis was performed on a Thermo Fisher Scientific applied Biosystems QuantStudio 5 system (Applied Biosystems) using the SYBR Green PCR Master Mix (Applied Biosystems). The reaction consisted of 10 μl of SYBR Green PCR Master Mix, 1 μl of a 5 μM mix of forward and reverse primers, 8 μl of water, and 1 μl of template cDNA in a total volume of 20 μl. Cycling was performed using the QuantStudio Design Analysis Software. The relative expression of each gene was normalized against 18S rRNA. The data are presented as the mean±s.d. The primers used are listed in Table S1.

For miRNA qRT-PCR analysis, the total RNA was extracted from 2.0×10⁵ cells with the RNeasy Mini Kit (QIAGEN), and 1 μg was reverse transcribed using a miRcute Enhanced miRNA cDNA First Strand Synthesis Kit (KR211, TIANGEN). Real-time PCR was performed in triplicate for each sample using the miRcute Enhanced miRNA SYBR Green PCR Kit (FP411, TIANGEN). The primers used were designed by TIANGEN. RNA RNU6 was used as the endogenous reference RNA to normalize the amount of template added. The data are presented as the mean±s.d.

Total DNA from cells was obtained using the QIAamp DNA Mini Kit (QIAGEN). Quantitative real-time PCR was performed to quantify the mtDNA content. Primers used for this experiment were specific to mt-tRNA (Leu) (UUR), and the 18S rRNA unique gene sequences (Table S1) were used for the nuclear genomes.

The ultra-structural analysis was performed as previously described (Li et al., 2014). The samples were examined using a HT7700 transmission electron microscope (Hitachi).

Cells were lysed on ice using a Nuclear and Cytoplasmic Extraction Reagent Kit (NE-PER) (Thermo Scientific). The samples were normalized for protein concentration using the BCA protein assay. Each sample (15–20 µg) was analysed using 10% SDS-PAGE (Invitrogen) and transferred to a PVDF membrane (Merck Millipore). The membranes were blocked in 5% BSA in TBST, and were incubated overnight at 4°C with the specific primary antibodies (Table S2). Glyceraldehyde-phosphate dehydrogenase (GAPDH) and β-tubulin were used as internal reference. The membranes were washed with TBST and incubated with IRDye-conjugated secondary antibodies (Table S2) for 1 h at room temperature. The membranes were scanned with the Odyssey detection system (Li-COR). Relative densities of proteins were quantitatively assessed using ImageJ. The data are presented as the mean±s.d.

The rate of ATP production from the two key energy pathways was detected simultaneously in live cells using the Agilent Seahorse XF Real-Time ATP rate assay (Seahorse Bioscience). Briefly, hASCs were plated in seeding medium in collagen I-coated 24-well XF^e24 cell culture microplates, with 20,000 or 25,000 (for differentiation) cells per well, at 37°C with 5% CO₂ for definitive endodermal differentiation. At the end of the definitive endodermal induction step (day 5), cells were washed twice with warmed DMEM assay medium (XF assay-modified DMEM supplemented with 10 mM XF glucose, 1 mM XF pyruvate, 2 mM XF glutamine, pH 7.4) and incubated at 37°C in a non-CO₂ incubator for 45–60 min prior to the assay to allow cells to pre-equilibrate with the assay medium. Before starting the XF assay, the medium was removed and replaced with fresh, warmed assay medium. The OCR and ECAR measurements were then assessed using a Seahorse XF Real-Time ATP Rate Assay Kit (Seahorse Bioscience), according to the manufacturer's instructions, and using 1.5 µM oligomycin and 0.5 µM rotenone/antimycin A. After measurement, the cell number per well was counted, and the OCR and ECAR measurements were normalized to the number of cells per well. The ATP production rate, including the glycoATP production rate and the mitoATP production rate, was calculated using the Agilent Seahorse XF Real Time ATP Rate Assay Report Generator (Seahorse Bioscience). All measurements were normalized, and the three measurements of the basal (starting) level of the cellular OCR of each well were averaged. Each sample was measured in four wells. The experiments were repeated three times, with cells from four different donors.

The production of ROS was detected using the DCFDA Cellular ROS Detection Assay Kit (ab113851, Abcam). Briefly, at the end of the definitive endodermal induction step, cells were re-plated in 96-well black- or clear-bottomed polystyrene microplates (Corning), with 25,000 cells per well at 37°C with 5% CO₂ and cultured overnight separately. The cells were then pre-incubated with 25 µM diluted DCFDA solution (Abcam) for 45 min at 37°C in the dark. Next, the cells were washed twice with 1×buffer (provided by DCFDA Cellular ROS Detection Assay Kit, Abcam), and 200 μM TBHP (Abcam) was added for 2 h at 37°C in the dark. The intracellular levels of ROS were immediately measured by a fluorescence plate reader (PerkinElmer), with excitation and emission spectra of 495 nm and 529 nm, respectively.

At least three independent determinations of each parameter were compared among the treatment groups by unpaired two-tailed Student’s t-test using the statistical software SPSS 11.5 (IBM Corporation). The differences were considered significant if P<0.05.

The authors thank Chenguang Zhang, Jun Deng, Man Ji and Hua Wei for advice in performing these experiments.