TWIST1 Silencing Enhances In Vitro and In Vivo Osteogenic Differentiation of Human Adipose-Derived Stem Cells by Triggering Activation of BMP-ERK/FGF Signaling and TAZ Upregulation

Abstract

Mesenchymal stem cells (MSCs) show promise for cellular therapy and regenerative medicine. Human adipose tissue-derived stem cells (hASCs) represent an attractive source of seed cells in bone regeneration. How to effectively improve osteogenic differentiation of hASCs in the bone tissue engineering has become a very important question with profound translational implications. Numerous regulatory pathways dominate osteogenic differentiation of hASCs involving transcriptional factors and signaling molecules. However, how these factors combine with each other to regulate hASCs osteogenic differentiation still remains to be illustrated. The highly conserved developmental proteins TWIST play key roles for transcriptional regulation in mesenchymal cell lineages. This study investigates TWIST1 function in hASCs osteogenesis. Our results show that TWIST1 shRNA silencing increased the osteogenic potential of hASCs in vitro and their skeletal regenerative ability when applied in vivo. We demonstrate that the increased osteogenic capacity observed with TWIST1 knockdown in hASCs is mediated through endogenous activation of BMP and ERK/FGF signaling leading, in turn, to upregulation of TAZ, a transcriptional modulator of MSCs differentiation along the osteoblast lineage. Inhibition either of BMP or ERK/FGF signaling suppressed TAZ upregulation and the enhanced osteogenesis in shTWIST1 hASCs. Cosilencing of both TWIST1 and TAZ abrogated the effect elicited by TWIST1 knockdown thus, identifying TAZ as a downstream mediator through which TWIST1 knockdown enhanced osteogenic differentiation in hASCs. Our functional study contributes to a better knowledge of molecular mechanisms governing the osteogenic ability of hASCs, and highlights TWIST1 as a potential target to facilitate in vivo bone healing.

Introduction

Regenerative medicine and tissue engineering both harness the potency of human cells to repair, regenerate, and even recreate tissues and organs with the goal of restoring their architecture and functionality. This emerging field requires a reliable source of stem cells in addition to biomaterial scaffolds and cytokine growth factors. As a potential cell source for tissue engineering, human adipose tissue-derived stem cells (hASCs) have attracted attention in this field [1, 2]. hASCs represent a promising tool for new clinical concepts in supporting cellular therapy because they are able to self-renew, and possess a mesodermal differentiation potential, with the ability to differentiate along multiple lineage pathways [3–5]. Moreover, use of hASCs circumvents the limited cell availability and painful harvest procedures necessary to obtain bone marrow-derived mesenchymal stem cells (MSCs) or bone grafts [6–9].

Among various mesoderm cell lineages, hASCs can also differentiate into the osteoblast lineage. The osteogenic capacity of ASCs has been studied extensively both in vitro and in vivo [4, 10–15]. Several in vivo studies using animal models support the potential translation for hASCs use in the treatment of human skeletal defects [13–16]. However, the repair of large bone defects due to trauma, inflammation, congenital defects, osteoporosis, or tumor represents yet a major challenge in reconstructive surgery. Therefore, current research efforts are focusing on how to optimize the use of hASCs and other mesenchymal cells for skeletal regenerative medicine.

Basic helix-loop-helix (bHLH) transcription factors play key roles in developmental processes such as lineage commitment and cellular differentiation [17, 18]. Twist1, an important member of bHLH transcription factors family, is an evolutionarily highly conserved protein [19] originally identified in Drosophila [19, 20] which shares extensive homology with Dermo-1 (Twist2), another bHLH transcription factor [21]. Early developmental studies in Drosophila showed that Twist is specifically expressed in mesodermal and cranial neural crest cells [22] and plays a key role in mesoderm formation and myogenesis through activation of downstream genes [19, 20, 23, 24]. For example, during mesoderm specification and differentiation, Twist has been implicated as an epithelial-mesenchymal transition regulator [25] for its role in activating DN-cadherin during Drosophila embryogenesis [26]. Other studies in sea urchin further support the notion that Twist is a crucial regulator of the skeletogenic gene regulatory network [27]. In vertebrates, Twist proteins exhibit bifunctional roles as both activators and repressors of gene transcription, and the mechanisms of their transcriptional activity are complex [28]. This transcription factor may act by first homo- or heterodimerizing with other bHLH molecules, and subsequently binding to a conserved motif known as the E box (CANNTG), which leads to either activation or inhibition of transcription [17, 29, 30].

Twist1 is expressed in the vertebrate skeletal mesenchyme and plays key roles in the control of mesenchymal cell lineage allocation during skeletal development as supported by several in vivo studies [31, 32]. During limb morphogenesis, Twist1 contributes in mediating cell patterning [33–35]. Several studies have indicated that Twist1 also plays a role in skull vault and craniofacial development [32, 36, 37]. Twist1 is expressed in vivo by osteoprogenitor cells but not by mature osteoblasts [32]. Moreover, it has been shown that Twist proteins transiently inhibit Runx2 function and expression during skeletogenesis [38, 39]. A study by Rice et al. demonstrated that lack of one Twist allele in mice caused misexpression of fibroblast growth factor receptor 2 (Fgfr2) in the sagittal suture [32]. In humans, haplo-insufficiency of TWIST1 is associated with the autosomal dominant Saethre-Chotzen syndrome [40–42], characterized by a varied pattern of craniofacial defects including craniosynostosis, a premature cranial suture fusion [40, 43] as a result of increased bone formation [44], indicating that Twist1 is an important transcription factor controlling osteoblastogenesis [18, 32, 42]. Recently, it has been reported that Twist1 controls a cell specification switch governing cell fate decisions within the cardiac neural crest, thereby regulating cell fate determination between ectodermal and mesodermal lineages [45].

Twist1 is abundantly and selectively expressed also in the adult adipose tissue, and its constitutive expression is higher in subcutaneous versus visceral fat in both mice and humans [46, 47]. Furthermore, recent studies indicate a novel role for Twist1 as a potential regulator of adipose tissue remodeling and inflammation [48].

In this study, we used a shRNA-based technique to investigate the contribution of TWIST1 to the osteogenic potential of hASCs. Our data indicate that TWIST1 silencing increases the osteogenic potential of hASCs in vitro and in vivo, through endogenous activation of BMP and ERK/FGF signaling, and upregulation of transcriptional coactivator TAZ. Moreover, we identify TAZ as downstream effector through which TWIST1 silencing triggers the osteogenic potential of hASCs. Collectively, our results indicate TWIST1 as a potential target to facilitate in vivo bone healing cell therapy-mediated.

Materials and Methods

Cell Harvest and Culture Conditions

All experiments followed protocols approved by the Animal Facility at Stanford University. All research involving human tissue has been approved by the Stanford Institutional Review Board, protocol #2188. All research involving vertebrate animals has been approved by Stanford APLAC, protocol #9999. Human ASCs were obtained from lipoaspirates of patients after informed consent, following approved guidelines by the Stanford University's Institutional Review Board. Primary cell culture of hASCs was performed as previously described [14, 49]. Briefly, lipoaspirates were washed with sterile phosphate buffered saline (PBS) (GIBCO, Invitrogen, Carlsbad, CA) and digested with 0.075% Collagenase I (Sigma-Aldrich, St. Louis, MO) for 60 minutes in a shaking water bath at 37°C. Collagenase was inactivated with Dulbecco's modified Eagle's medium (DMEM)-GlutaMAX (GIBCO, Invitrogen, Carlsbad, CA) medium containing 10% fetal bovine serum (FBS) (GIBCO, Invitrogen, Carlsbad, CA) and subsequently centrifuged at 1,200g. The pellet was resuspended in DMEM-GlutaMAX medium and filtered through a 100 μm nylon mesh to remove debris. The filtrate was recentrifuged, cells were plated on PRIMARIA tissue culture dishes (BD Falcon, Frankline Lakes, NJ), and cultured in DMEM-GlutaMAX medium containing 10% FBS and 100 IU/ ml Penicillin/Streptomycin (GIBCO, Invitrogen, Carlsbad, CA) in incubators at 37°C and 5% atmospheric CO2. Cells were passaged upon reaching subconfluence via a 5 minutes 37°C incubation in 0.25% trypsin (GIBCO, Invitrogen, Carlsbad, CA). Only first and second passage cells were used for all experiments.

RNA Isolation, Reverse-Transcriptase Polymerase Chain Reaction, and Quantitative PCR Analysis

Total RNA was isolated from cells by the TRIzol method (Invitrogen). Purified and quantified RNA was treated with DNase I (Ambion, Austin, TX) to clear genomic DNA. Reverse transcription (RT), PCR, and quantitative real-time PCR (qPCR) were described previously [49–51]. Briefly, qPCR was performed using the ABI Prism 7900 Sequence Detection System, TaqMan Gene Expression Master Mix, and TaqMan Gene Expression Assays (Applied Biosystems, Grand Island, NY). PCR reactions were carried out under the following conditions: 94°C for 5 minutes, 94°C for 30 seconds, 55°C/60.3°C for 1 minute, and 72°C for 1 minute (25–30 cycles). Primers for RUNX2, ALPL, BGLAP, and GAPDH were previously described [52, 53]. Specific primers for other genes examined were designed based on their GenBank sequence, their sequences are listed in Table 1. The relative mRNA level in each sample was normalized to its GAPDH content. Values are given as relative to GAPDH expression. The results are presented as means ±SD of triplicate.

Table 1. Primers sequence.

Gene	Forward primer	Reverse primer	Accession #number
TWIST1	TCTTACGAGGAGCTGCAGACGCA	ATCTTGGAGTCCAGCTCGCGCT	NM_000474
TA Z	CTTGGATGTAGCCATGACCTT	TCAATCAAAACCAGGCAATG	NM_181311.2
FGFR2	TGGAGCGATCGCCTCACCG	CTTCCAGGCGCTGGCAGAACTGT	NM_000141
BMPRIB	TACAAGCCTGCCATAAGTGAAGAAGC	ATCATCGTGAAACAATATCCGTCTG	NM_001203
BMP-2	GGAAGAACTACCAGAAACGAG	AGATGATCAGCCAGAGGAAAA	NM-001200
PPARƔ2	CTCCTATTGACCCAGAAAGC	TCAAAGGAGTGGGAGTGGTC	U63415
ADIPSIN	GACACCATCGACCACGAC	CCACGTCGCAGAGAGTTC	NM_001928
OSX	CCCCACCTCTTGCAACCA	CCTTCTAGCTGCCCACTATTTCC	AF47798

TWIST1 and TAZ Silencing in hASCs Using shRNA Method

Knockdown of TWIST1 in hASCs was achieved with TWIST1 shRNA lentiviral particles according to the manufacturer's protocol (Santa Cruz Biotechnology, Santa Cruz, CA). Briefly, hASCs were plated onto 12-well plates 1 day before transduction to 50% confluence. Cells were transduced with either 12 ml of TWIST1shRNA lentiviral particles (sc-38604-V; Santa Cruz Biotechnology, Santa Cruz, CA) or control shRNA (scramble) lentiviral particles-A (sc-108080; Santa Cruz Biotechnology, Santa Cruz, CA) per well in DMEM-GlutaMAX medium plus 5 mg/ml polybrene for 6 hours. Thirty 12 wells at time, for each different viral particles (shTWIST1; shTAZ; double shTWIST1/ TAZ; shscramble) were used. Each set of transduction was carried out several time (at least five times), in order to perform all in vitro and in vivo experiments using only first and second passages cells. The efficiency of transduction monitored 24 hours after infection using green fluorescent protein control lentiviral particles (sc-108084, Santa Cruz Biotechnology, Santa Cruz, CA) was found to be 85%–87%. Two days after transduction, the cells underwent to puromycin selection (5 μg/ml) for 5 days. After puromycin treatment, no viable cells were observed in the multiwell containing mock transduced cells. The transduced cells, puromycin resistant, were collected by trypsin procedure. Approximately 300,000/35,000 cells were obtained from each 12-multiwell. Only first and second passages were used for all experiments. Knockdown of TAZ was performed as above using TAZ shRNA lentiviral particles (sc-38568-V; Santa Cruz Biotechnology, Santa Cruz, CA). Efficiency of TWIST1 and TAZ knockdown was assessed at gene expression level by RT-PCR analysis and at protein level by immunoblotting analysis. Cells transduced with TWIST1-shRNA, TAZshRNA, and scramble shRNA lentiviral particles are referred as to shTWIST1 hASCs, shTAZ hASCs, and scramble hASCs, respectively. Cells cotransduced with TWIST1 and TAZ shRNA lentiviral particles are referred as to shTWIST1/TAZ hASCs.

Induction of Osteogenic and Adipogenic Differentiation

Cells were seeded at a density of 80,000 per well in a six-well plate and allowed to grow in DMEM-GlutaMAX and 10% FBS for 2–3 days until confluence was reached. Osteogenic differentiation of hASCs was induced by culturing cells in DMEM-GlutaMAX supplemented with 10% FBS, 100 IU/ml penicillin, 100 IU/ml streptomycin, 10 mM β-glycerophosphate, and 100 μg/ml ascorbic acid (Sigma-Aldrich, St Louis, MO) (Osteogenic Differentiation Medium, ODM). Osteogenic differentiation in presence of BMP-2 and ERK1/2-FGF signaling inhibitors was performed by adding to the osteogenic medium 250 ng/ml of recombinant human noggin protein (R&D Systems, Minneapolis, MN) or 5 mM U-0126 (Sigma-Aldrich, St Louis, MO), respectively. The optimal inhibitors concentration was determined based on previous dose-response studies. The used concentrations did not affect cell survival. Medium was changed every 3 days. Adipogenic differentiation was induced by culturing cells for 3 days in adipogenic differentiation medium (DMEM 10% FBS, 1% penicillin/streptomycin, 10 μg/ ml insulin, 1 μM dexamethasone, 0.5 mM methylxanthine, and 200 μM indomethacin), which was then changed to adipocyte maintenance medium (DMEM, 10% FBS, 1% penicillin/ streptomycin, 1 μg/ml insulin) for an additional 3 days before assessment of adipogenic differentiation. After induction, cultures were fixed with 10% buffered-formaline phosphate, then stained with 0.1% oil red O solution.

Alkaline Phosphatase Activity and Mineralization Assay

Alkaline phosphatase (ALPL) activity, Alizarin red, and von Kossa mineralization staining were performed as previously described [51, 52]. Briefly, alkaline phosphatase activity was determined by biochemical colorimetric assay using Alkaline Phosphatase kit #104-LS, (Sigma-Aldrich, St Louis, MO) according to the manufacturer's instructions. After seeding cells (80,000 per well in a six-well plate) and growing them to confluence, osteogenic differentiation was initiated. After 10 days of osteogenic differentiation, ALPL activity was determined in cell lysates by measuring levels of p-nitrophenol, a metabolite formed during hydrolysis, via optical density measurements read at 420 nm. Values were normalized against protein concentration using a bicinchoninic acid (BCA) protein assay kit (Pierce Biotechnology, Rockford, IL). Samples were run in triplicate. For Alizarin red staining, cells were washed with PBS twice, fixed with 70% ethanol at 4°C for 30 minutes, and then washed with deionized water. Cells were stained with 0.2% Alizarin red for 1 hour at room temperature, then washed with PBS three times. Cell culture plates were air-dried and evaluated by light microscopy using an inverted microscope. Images were acquired using a ScanJet 5370C scanner (Hewlett-Packard Company, Palo Alto, CA). Alizarin red staining was quantified by colorimetric assay, incubating stained cells with a 20% methanol 10% acetic acid solution for 30 minutes to elute all calcium-bound stain. Optical density was determined at 450 nm. Values were normalized against protein concentrations obtained from triplicate wells. For von Kossa staining, cells were washed twice with PBS, fixed with 70% ethanol for 30 minutes at room temperature, and rinsed with deionized water twice. Cells were stained with freshly made 5% Silver Nitrate for 1 hour under UV then rinsed with deionized water and incubated for 2 minutes in 5% sodium thiosulfate followed by washing with deionized water. Images were acquired as described above.

Immunoblotting Analysis

Immunoblotting analysis was performed using the following primary rabbit antibodies: anti-pSMAD1/5 (Ser463/465), anti-SMAD5, anti-pERK1/2 (Thr202/Tyr204) (dilution 1:1,000; Cell Signaling Danvers, MA), anti-ERK1/2 (C-14: sc-154), anti-TWIST1 (H-8: sc-15393), anti-TAZ (H-70: sc-48805), anti-FGFR2/Bek (H-80: sc-20735), anti-BMPRIB (H-44: sc-25455) (dilution 1:200, Santa Cruz Biotechnology, Inc.), and anti-β-ACTIN (ab8227) (dilution 1:5,000; Abcam, Cambridge, MA). Fifty micrograms of cell lysate protein isolated from cells was resolved by NuPAGE 4%–12% bis-Tris-HCl sodium dodecyl sulfate-polyacrylamide gel (Novex, Life Technologies, Carlsbad, CA). Proteins were transferred to a polyvinylidene fluoride membrane (Bio-Rad, Inc., Hercules, CA). Membranes were probed with specific antibody. Horseradish peroxidase-conjugated secondary anti-rabbit antibody was used (dilution 1:2,000; Cell Signaling, Danvers, MA). Immunoblotted proteins were visualized by enhanced chemiluminescence (Amersham Biosciences, Buckinghamshire, U.K.). To assess for the total amount of endogenous SMAD5 or ERK1/2, and to control for equal loading and transfer of the samples, the membranes were stripped and reprobed with either anti-SMAD5 or anti-ERK1/2 antibodies, and anti-β-ACTIN antibody. Densitometry analysis of electrophoretic bands was performed using the ImageJ software program (NIH, Bethesda, MA). The density of each phosphorylated band was normalized to the loading controls (β-ACTIN) and presented as percentage increase or decrease. The results are the mean±SD of three independent experiments.

Immunofluorescence Staining

Aliquots of same cells used for in vivo experiments were seeded on Lab-Tek II chamber slide (100,000 cells/well) (Nalge Nunc, Thermo Scientific, Rochester, NY). After 8 hours, cells were fixed with methanol for 5 minutes at −20°C then with acetone for 2 minutes at −20°C. After washing five times with cold PBS, cells were incubated with 2% normal goat serum/PBS for 1 hour at room temperature to block nonspecific binding of antibodies. Subsequentially, the cells were incubated either with rabbit primary anti-phosphoS-MAD1/5 antibody or antiphosphoERK1/2 (dilution 1:50; Cell Signaling, Danvers, MA) overnight at 4°C followed by fluorescein-conjugated Alexa Fluor 568 anti-rabbit secondary antibody (dilution 1:400; Molecular Probes, Invitrogen, Carlsbad, CA) for 1 hour at room temperature. Nuclear counter-staining was performed using Vectashield H-1200 mounting medium with DAPI (Vector Laboratories, Burlingame, CA). A Zeiss Axioplan-2 microscope equipped with Axiocam HRc digital camera (Zeiss, Thornwood, NY) was used for imaging.

Preparation of Cell-Conditioned Media and BMP-2 Enzyme-Linked Immunosorbent Assay

Cell-media were collected at indicated time points from shTWIST1 hASCs and scramble hASCs undergoing osteogenic differentiation in ODM supplemented with 2% foetal calf serum (FCS).Then, media were concentrated 50-fold using Centricon filters (3000 NMWL, Millipore Corporation, Billerica, MA). Collection and concentration of the media were carried out at 4°C. The volume of each conditioned medium was normalized by cell numbers so that an equal volume of shTWIST1 and scramble medium was produced by an equal number of shTWIST1 and scramble cells. Protein concentration was determined by BCA protein assay (Pierce Biotechnology, Rockford, IL). Media were analyzed for BMP-2 concentrations by enzyme-linked immunosorbent assay (ELISA) using Quantakine human BMP-2 kit #DB100B (R&D Systems, Minneapolis, MN), according to the manufacturer's instructions. Photometric detection was done with an ELISA reader at 370-nm wavelength. Each sample was run in triplicate.

Statistical Analysis

Data are expressed as mean ± SD of at least three independent samples. Statistical comparisons between groups were performed with a two-tailed Student's t test, *, p≤ .05; **, p≤ .01; and ***, p≤ .001 were considered significant.

Cell Seeding on Scaffolds

Scramble and TWIST1sh hASCs were trypsinized, washed with FBS, and counted. Cells (250,000) were suspended in 100 μl FCS and seeded on apatite-coated poly(lactic-co-glycolic acid) (PLGA) scaffolds which were fabricated from 85/15 poly(lactic-co-glycolic acid) by solvent casting and a particulate leaching process as previously described [13]. In order to best mimic potential clinical translational scenarios, scaffolds were implanted in calvarial defects 2 hours after in vitro seeding.

Animal Surgery

All animal experiments were performed in accordance with Stanford University Animal Care and Use Committee guidelines. For evaluating the in vivo healing capacity of TWIST1sh hASCs 21-day-old male nude CD1-mice (Charles River Laboratories, Wilmington, MA) underwent calvarial defect procedures as previously described [14, 53]. Briefly, after anesthesia with an intraperitoneal injection of ketamine 100 mg/kg + xylazine 20 mg/kg + acepromazine 3 mg/kg and disinfection of the surgical site of the mice, nonhealing critical 4-mm calvarial defects were created with a trephine drill bit in left parietal bones as previously described [14]. Meticulous care was taken in order to protect the underlying dura mater or neighboring cranial sutures. Treatment groups included no treatment (empty), scaffold with serum, scaffold seeded with shTWIST1 hASCs, scaffold seeded with scramble hASCs, scaffold seeded with shTWIST1/TAZ hASCs, and scaffold seeded with shTAZ hASCs. The scaffolds were placed in the defects, the wound was closed, and the animals were allowed to recover.

μCT-Scanning

μCT-scanning was performed as previously described [14]. Briefly, CD1 nude mice were scanned with a high-resolution MicroCAT II scanner (ImTek, Inc., Knoxville, TN) with an x-ray voltage of 80 kVP and an anode current of 450 μA. A resolution of 80 μm was obtained with 144 steps over 360 Åã rotation. X-ray data reconstruction was performed with Cobra EXXIM (EXXIM Computing, Corp., Livermore, CA), and Micro View Software (GE Healthcare, Buckinghamshire, UK). Each mouse was scanned with a CT-phantom, which was used to calibrate each scan. The precise threshold for regenerating calvarial bone was previously determined equivalent to 510 Houndsfield Units. The rest-defect area was then determined with the Magic Wand Tool in Photoshop (Adobe Systems, San Jose, CA). Percentage healing was determined by dividing the rest-defect area by the mean of the defect size 1 day postoperatively. CD1 nude mice were scanned 24 hours postsurgery and at weeks 2, 4, 6, and 8. For statistical analysis, the shTWIST1 hASCs group was compared with all other groups using the Mann-Whitney U test. A *, p-value< .05 was considered statistically significant.

Histology

Skulls were harvested under a stereomicroscope and fixed in 10% neutral buffered formalin overnight at 4°C and decalcified in 19% EDTA for the appropriate time. Samples were then dehydrated and paraffin embedded. Pentachrome staining was performed on 8 μm coronal sections according to standard procedures. Sections were examined with a Carl Zeiss Axioplan-2 microscope (Zeiss). Images were captured by AxioVision HRc camera (Zeiss) and combined by Adobe Photoshop (Adobe Systems).

Results

Endogenous Expression of TWIST1 Is Downregulated in hASCs Undergoing to Osteogenic Differentiation

We began our study by analyzing the endogenous expression profile of TWIST1 in undifferentiated hASCs as well as during their differentiation along the osteogenic lineage. At day 0, quantitative PCR (qPCR) analysis revealed high expression of TWIST1 in undifferentiated hASCs (Fig. 1A). Interestingly, TWIST1 expression was sharply downregulated in hASCs undergoing to osteogenesis (Fig. 1A). A similar pattern was also observed at protein level as indicated by immunoblotting analysis of TWIST1 protein and quantification of electrophoresis bands by Image J software program (Fig. 1B, 1C). To verify whether downregulation of TWIST1 expression occurred only in differentiating hASCs and that it was not merely an out-coming of culture timing, we examined TWIST1 expression profile in hASCs cultured in growth medium for the same length of time as for the osteogenic differentiation assay. The results obtained from this analysis did not show overtime, significant differences in the expression of endogenous TWIST1 in hASCs maintained in growth medium (Fig. 1A–1C). Therefore, TWIST1 downregulation was peculiar to hASCs differentiating into osteoblasts.

Figure 1.

Endogenous TWIST1 is downregulated during osteogenesis of human adipose tissue-derived stem cells (hASCs). (A): qPCR analysis performed on hASCs undergoing to osteogenesis reveals a significant downregulation of TWIST1 gene expression overtime. Conversely, TWIST1 expression remains steady in hASCs cultured in growth medium for the same length of time. (B): Immunoblotting of TWIST1 proteins performed using specific anti-TWIST1 antibody confirms the patterns observed at gene expression level. Membranes were stripped and reprobed with β-ACTIN antibody as loading control. (C): Histograms representing quantification of TWIST1 protein obtained by Image J program. (D): Gene expression analysis of TWIST1 and osteogenic markers RUNX2, OSX, and BGLAP examined by qPCR reveals that TWIST1 expression inversely correlates with that of early osteogenic markers RUNX2 and OSX as well late marker osteocalcin (BGLAP). The relative mRNA level in each sample is normalized to its GAPDH content. Values are given as relative to GAPDH expression. *,p≤.05; **, p ≤.01; and ***, p ≤.001. Abbreviations: GM, growth medium; ODM, osteogenic medium.

Interestingly, decreased expression of TWIST1 was paralleled by upregulation of early osteogenic marker RUNX2, followed by upregulation of late osteogenic marker osteocalcin (BLGP) at day 21 (Fig. 1D). Thus, our analysis demonstrated that TWIST1 expression inversely correlated with the expression of osteogenic markers in hASCs differentiating along the osteogenic lineage.

Knockdown of TWIST1 Expression in hASCs by shRNA

To examine potential function of TWIST1 on the osteogenic potential of hASCs, we knocked-down endogenous TWIST1 expression in hASCs, using shRNA technology. Therefore, cells were transduced with a pool of lentiviral particles containing three target-specific constructs encoding plus hairpin shRNA human TWIST1. Control hASCs were transduced with scramble shRNA lentiviral particles. Upon puromycin selection, in stable transduced hASCs TWIST1 resulted in a 70% decrease at mRNA level as indicated by RT-PCR and its quantification (Fig. 2A, 2B, top panel). Decreased TWIST1 was also confirmed at protein level by immunoblotting analysis (Fig. 2A, 2B, lower panel).

Figure 2.

Knockdown of TWIST1 enhances in vitro osteogenic potential of human adipose tissue-derived stem cells (hASCs). (A): Validation of TWIST1 knockdown by shRNA technique assessed by RT-PCR and immunoblotting analysis. (B): Histograms showing quantification of TWIST1 gene expression (top panel), and TWIST1 protein (bottom panel) obtained by Image J program. (C): Alizarin red staining at day 21 of osteogenic differentiation reveals a robust mineralization of extracellular matrix in shTWIST1 hASCs as compared to scramble hASCs. (i) high magnification (×200) and (ii) lower magnification (×40). (D): A prominent matrix mineralization and larger bone nodules are also detected by von Kossa staining in shTWIST1 hASCs compared to scramble hASCs at day 28. Magnifications as above. (E): Expression of osteogenic marker genes by qPCR analysis showing significant upregulation of early markers RUNX2 and OSX, and late marker BGLAP in shTWIST1 hASCs as compared to scramble cells. The relative mRNA level in each sample is normalized to its GAPDH content. Values are given as relative to GAPDH expression. *, p≤.05 and **, p≤.01.

Downregulation of Endogenous TWIST1 Increases the In Vitro Osteogenic Potential of hASCs

Next, we sought to investigate the impact of knocking-down TWIST1 on the osteogenic ability of hASCs. Therefore, to test how shTWIST1 hASCs differentiate into functional osteoblasts capable of producing mineral deposits, cells were grown in osteogenic differentiation media for 21/28 days and then stained with either 0.2% Alizarin red S or 2% silver nitrate (von Kossa staining), which stain for calcium deposits. As shown in Figure 2C, 2D, shTWIST1 hASCs stained positive for Alizarin red S and von Kossa more intensely than scramble hASCs. These findings indicated that shTWIST1 hASCs produced and incorporated higher amounts of calcium into their extracellular matrix. In addition, molecular analysis of osteogenic markers at different time points of osteogenic differentiation assay showed significant higher level of RUNX2 at day 3 and followed by Osterix (OSX) at day 10 in shTWIST1 hASCs as compared to scramble hASCs (Fig. 2E). At late time point of osteogenic assay, shTWIST1 hASCs expressed also significantly more BLGP than scramble hASCs (Fig. 2E). Taken together, the above results indicated that TWIST1 downregulation increased the osteogenic potential of hASCs.

Downregulation of Endogenous TWIST1 Enhances BMP and ERK1/2-FGF Signaling Pathways in hASCs

It is known that differentiation of mesenchymal cells, such as ASCs, into osteoblasts is regulated by the BMP-mediated signaling pathway [54, 55]. As an attempt to unveil the molecular mechanism responsible for the increased osteogenic capacity of shTWIST1 hASCs, we investigated the degree of endogenous activation of BMP signaling in these cells. Immunoblotting analysis of the downstream effectors of BMP signaling phosphoSMAD1/5 revealed increased levels of phosphoSMAD1/5 proteins in shTWIST1 hASCs as compared to control cells (Fig. 3A, 3B). Additionally, qPCR and immunoblotting analysis indicated that TWIST1 silencing increased BMP receptor-type IB (BMPRIB), both at gene and protein level in hASCs (Fig. 3C, 3D). Several reports have suggested that BMPRIB is critical to the osteogenic commitment of a cell with multilineage potential as hASCs [54, 56]. In addition, shTWIST1 hASCs undergoing to osteoblast differentiation elaborated significant more BMP-2 than scramble hASCs (Supporting Information Fig. S1).

Figure 3.

Downregulation of endogenous TWIST1 enhances BMP and ERK1/2-FGF signaling in human adipose tissue-derived stem cells (hASCs). (A): Extent of endogenous SMAD1/5 phosphorylation measured by blotting cell lysates with specific anti-pSMAD1/5 antibody reveals a remarkable and sustained activation of BMP signaling in shTWIST1 hASCs ongoing to osteogenic differentiation as compared to their scramble controls. Membranes were stripped and reprobed with β-ACTIN antibody to control for equal loading. SMAD1/5 proteins were detected by specific anti-SMAD1/5 antibody. (B): Histogram represents quantification of phosphorylated SMAD1/5 proteins obtained by Image J software program. (C): qPCR reveals increased expression of BMP receptor type IB (BMPRIB) in shTWIST1 hASCs as compared to scramble control. The relative mRNA level in each sample is normalized to its GAPDH content. Values are given as relative to GAPDH expression. (D): Immunoblotting analysis shows also higher level of BMPRIB protein in shTWIST1 hASCs. (E): Immunoblotting analysis of phospho-ERK1/2 with specific anti-phosphoERK1/2 antibody showing increased activation of ERK1/2 signaling in shTWIST1 hASCs. Membranes were stripped and reprobed with β-ACTIN antibody to control for equal loading. ERK1/2 proteins were detected by specific anti-ERK1/2 antibody. (F): Histogram is the quantification of phosphorylated ERK1/2 proteins obtained by Image J program. (G): qPCR indicates that shTWIST1 hASCs express higher amount of FGF receptor type 2 (FGFR2) as compared to scramble control. The relative mRNA level in each sample is normalized to its GAPDH content. Values are given as relative to GAPDH expression. (H): Immunoblotting analysis confirms increased FGFR2 at protein level in shTWIST1 hASCs. *, p≤.05; **, p≤.01; and ***, p≤.001.

Our investigation showed also enhanced activation of pERK1/2 signaling in shTWIST1 hASCs. As indicated by immunoblotting analysis and its quantification, higher amounts of phosphoERK1/2 proteins were present in shTWIST1 hASCs compared to control cells (Fig. 3E, 3F). ERK1/2 is another important signaling pathway involved in the control of osteoblastogenesis. It is well established that ERK1/2 is the signaling through which FGF promotes osteogenic differentiation of mesenchymal cells [57, 58]. Interestingly, we observed increased FGF receptor-type 2 (FGFR2) at gene and protein level in shTWIST1 hASCs (Fig. 3G, 3H). Thus, the above results suggested that downregulation of endogenous TWIST1 in hASCs leads to activation of two key osteogenic signaling pathways.

Upregulation of Endogenous TAZ in shTWIST1 hASCs Contributes to Their Enhanced Osteogenic Ability

TAZ a transcriptional coactivator with PDZ-binding motif drives MSCs to differentiate into osteoblast lineage, while simultaneous blocking their differentiation into fat [59, 60]. Several studies have suggested that both BMP and ERK1/2-FGF signaling pathways stimulate TAZ expression [60, 61]. Having gained evidence of enhanced active BMP and ERK1/2-FGF signaling in shTWIST1 hASCs we sought therefore, to investigate the endogenous profile of TAZ expression in these cells. Analysis by qPCR showed significant upregulation of TAZ expression in shTWIST1 hASCs compared to scramble controls throughout osteogenic differentiation (Fig. 4A). TAZ expression was already markedly increased in shTWIST1 hASCs at day 0, the expression level remained sustained at days 3 and 10. By day 21, when cells were fully differentiated to osteoblasts TAZ expression decreased (Fig. 4A). Immunoblotting analysis with specific anti-TAZ antibody revealed also higher TAZ protein in shTWIST1 hASCs (Supporting Information Fig. S2).

Figure 4.

Upregulation of TAZ in shTWIST1 human adipose tissue-derived stem cells (hASCs) participates to enhance the osteogenic potential. (A): Time-course qPCR analysis showing a significant upregulation of TAZ expression in shTWIST1 hASCs compared to scramble cells throughout osteogenic differentiation. (B): Validation of double TWIST1 and TAZ genes knockdown in hASCs (shTWIST1/TAZ). Lane 1, minus RT; lanes 2 and 3, endogenous expression of TWIST1 and TAZ in scramble controls; lane 4, downregulation of TWIST1 in shTWIST1 hASCs; lane 5, downregulation of TAZ in shTAZ hASCs; lane 6, downregulation of TWIST1 in double shTWIST1/TAZ hASCs; lane 7, downregulation of TAZ in double shRNA TWIST1/TAZ hASCs. (C): Immunoblotting analysis validates at protein level the effective knockdown of both TWIST1 and TAZ genes. Lane 1, endogenous TWIST1 protein in scramble control; lane 2, TWIST1 protein in shTWIST1 hASCs; lane 3, TWIST1 protein in double shRNA TWIST1/TAZ hASCs; lane 4, TAZ protein in double shRNA TWIST1/TAZ hASCs; lane 5, TAZ protein in shTAZ hASCs; lane 6, TAZ protein in scramble control. Histogram below represents quantification of TWIST1 and TAZ immuno-blotted proteins by Image J analysis. (D): Alizarin red staining showing a markedly decreased of extracellular matrix mineralization in double shTWIST1/TAZ hASCs compared to shTWIST1 hASCs. (E): Quantification of Alizarin red staining. (F): Alkaline phosphatase enzymatic (ALPL) activity determined at day 10, as an intermediate marker of osteogenic differentiation, shows significantly less enzymatic activity in double shTWIST1/TAZ hASCs compared to shTWIST1 hASCs. (G, H): Time course of RUNX2 and BGLAP gene expression by qPCR further confirms at molecular level the decreased osteogenic capacity of double shTWIST1/TAZ hASCs compared to shTWIST1 hASCs. The relative mRNA level in each sample is normalized to its GAPDH content. Values are given as relative to GAPDH expression. *, p≤.05; **, p≤.01; and ***, p≤.001. Abbreviation: RT, reverse transcriptase.

Next, we tested whether TAZ upregulation would, in part, be responsible for the greater osteogenic capacity of shTWIST1 hASCs. To this end, hASCs were cotransduced with two pools of lentiviral particles each containing three target-specific constructs encoding plus hairpin shRNATWIST1 and TAZ. Puromycin resistant clones were selected and analyzed at mRNA and protein level for both, TWIST1 and TAZ. Cells showing similar degree of TWIST1 and TAZ downregulation (Fig. 4B, 4C and lower panel) were subse-quentially tested for the osteogenic differentiation ability. In addition, shTAZ hASCs, shTWIST1 hASCs, and scramble controls were included in the assay. As shown in Figure 4D–4H, silencing of TAZ in shTWIST1 hASCs suppressed the increased osteogenic differentiation previously observed in these cells. This dramatic effect was demonstrated by reduced Alizarin red staining, reflecting poor mineralization of extracellular matrix, (Fig. 4D, 4E) less ALPL enzymatic activity (Fig. 4F), as well decreased expression of osteogenic markers RUNX2 and BGLAP (Fig. 4G, 4H) in double shRNA shTWIST1/TAZ hASCs. Furthermore, silencing of TAZ alone partially impaired osteogenic differentiation of hASCs as compared to scramble hASCs (Fig. 4D–4H).

Inhibition of BMP and ERK1/2-FGF Signaling Abolishes TAZ Upregulation and the Enhanced Osteogenic Ability in shTWIST1 hASCs

Our previous observation that the enhanced osteogenic potential of shTWIST1 hASCs correlated with more active endogenous BMP and ERK1/2-FGF signaling and TAZ upregulation, together with the knowledge that active BMP and ERK1/2-FGF signaling induced TAZ expression [60, 61], prompted us to hypothesize that inhibition of BMP and ERK1/2-FGF signaling would abrogate TAZ upregulation and the greater osteogenic capacity in shTWIST1 hASCs. To verify these hypothesis, two sets of osteogenic differentiation assays were performed in presence either of noggin, an inhibitor of BMP signaling or U0126 which inhibits the MAPK/ERK-mediated FGF signaling [57, 62]. The effective inhibition of BMP and ERK1/2-FGF signaling, as assessed by immunoblotting analysis (Fig. 5A), significantly decreased the level of TAZ expression in shTWIST1 hASCs (Fig. 5B). Importantly, both inhibitors downregulated TAZ expression and osteogenic differentiation in a dose-dependent fashion (data not shown). Moreover, inhibition of BMP and ERK1/2-FGF signaling impaired dramatically the osteogenic differentiation of shTWIST1 hASCs. We observed significant differences in ALPL enzymatic activity (Fig. 5C), mineralization of extracellular matrix (Fig. 5D, 5E), and expression of osteogenic markers RUNX2 and BGLAP (Fig. 5F) between untreated and inhibitors-treated shTWIST1 hASCs, with the latter performing less osteogenic differentiation. Taken together, these results suggested that BMP and ERK1/2-mediated signaling together participate to upregulate TAZ expression and enhance osteogenic differentiation induced by TWIST1 silencing in hASCs.

Figure 5.

Inhibition of BMP and ERK1/2-FGF signaling pathways abrogates TAZ upregulation and the enhanced osteogenic differentiation in shTWIST1 human adipose tissue-derived stem cells (hASCs). (A): Immunobloting analysis shows the effective inhibition of BMP and ERK/ FGF signaling as indicated by reduced levels of pSMAD1/5 and pERK1/2 proteins in cells treated with inhibitors. (B): Time course of TAZ gene expression by qPCR analysis on shTWIST1 hASCs treated either with noggin (250 ng/ml), inhibitor of BMP signaling, or U-01269 (5 μM), inhibitor of MAPK/ERK, mediator of FGF signaling. Treatments with both inhibitors significantly decreased the expression of TAZ compared to untreated cells. (C): At day 10 of osteogenic differentiation assay, ALPL enzymatic assay reveals significantly less activity in treated cells as compared to untreated control. (D): Poor osteogenic capacity of noggin and U-0126-treated shTWIST1 hASCs is also confirmed by Alizarin red staining performed at day 21. (E): Quantification of Alizarin red staining. (F, G): qPCR analysis of osteogenic markers RUNX2 and BGLAP showing significant reduced expression of both markers in treated cells. The relative mRNA level in each sample is normalized to its GAPDH content. Values are given as relative to GAPDH expression. *,p≤.05; **, p≤.01; and ***, p≤.001.

shTWIST1 hASCs Accelerate and Potentiate Bone Repair In Vivo

To adequately translate in vitro findings to the clinical realm, compelling in vivo data must be obtained showing that knocking-down TWIST1 enhances the skeletal repair capacity of hASCs. Therefore, we compared in vivo the ability of shTWIST1 hASCs to “scramble” hASCs group to heal calvarial defects. To this end, nonhealing critical sized (4 mm) calvarial defects were created in the parietal bone of nude mice as previously described [14]. The calvarial defects were treated with shTWIST1 hASCs and “scramble” hASCs loaded on PLGA scaffold [13, 14] (Fig. 6A, 6B). Serum-soaked scaffold and empty defects were used as controls. Prior loading on scaffolds, fractions of shTWIST1 hASCs and scramble hASCs were analyzed for endogenous levels of pSMAD1/5 and ERK1/2 by indirect immunofluorescence assay to ensure the occurrence of enhanced BMP and ERK1/2 signaling in shTWIST1 hASCs (Fig. 6A).

Figure 6.

shTWIST1 human adipose tissue-derived stem cells (hASCs) accelerate and potentiate bone repair in vivo. (A): Indirect immunofluorescence performed on aliquots of shTWIST1 hASCs and scramble control prior loading onto scaffolds, as described under Materials and Methods section, show increased levels of endogenous pSMAD1/5 and pERK1/2 in shTWIST1 hASCs compared to scramble hASCs. Scale bar = 50 μm. (B): Schematic representation of parietal calvarial defect (“critical”/nonhealing) treatment on 21-day-old nude mice skull. Four-millimeter nonhealing (4 mm) defects were made in the parietal bone of 21-day-old nude mice. (C): Quantification of defect repair according to μCT-scan results. Statistical analysis was conducted using the Mann-Whitney U test. p-Values: *, p≤.05. (D): Time-course of three representative mouse-treated μCT-scans. (E): Pentachrome staining of coronal sections of skull at postoperative week 10 showing the repair of calvarial bone defects as determined by yellow color. Scale bar = 400 μm (objective magnification ×25). Abbreviation: PLGA, poly(lactic-co-glycolic acid).

In order to investigate the ability of shTWIST1 hASCs to promote bone formation in vivo, calvarial healing was first monitored by mCT-analysis at different time points (0, 2, 4, 6, and 8 weeks) (Fig. 6C, 6D). Already at postoperative week 2, high healing rates were observed in shTWIST1 hASCs (11%) treated defects (n = 5 7), whereas the healing rate of scramble-treated defects (n = 7) was lower (3.8%).

Healing in the control groups was marginal (below 1.2%). This trend continued to week 8, when healing was found to be (55%) in shTWIST1 hASCs. Minor healing was observed in serum-soaked scaffold (n = 5) and empty defect control groups (n = 5) (less than 10%). In a direct comparison between the different groups, the healing capacity of shTWIST1 hASCs-treated defects was significantly better at weeks 4, 6, 8, as compared to scramble-treated defects. Differences between shTWIST1 hASCs and scramble-treated defects were likewise highly significant. No significant differences were found between serum-soaked scaffold and empty-treated defects at all-time points. Remarkably, as previously demonstrated in vitro (Fig. 4D–4H), cosilencing of TAZ (n = 4) in vivo suppressed the enhanced skeletal repair ability of shTWIST1hASCs (Fig. 6C, 6D) thus, confirming TAZ as a key mediator. Of note, the healing rate of shTWIST1/TAZ hASCs-treated defects mirrored that of scramble hASCs-treated defects. Finally, shTAZ alone partially affected healing defects compared to scramble-treated groups.

To assess for bone regeneration at histological levels, coronal sections of skull with bone defect were stained with Pentachrome procedure. Pentachrome staining of defects, harvested 10 weeks after surgery, confirmed the superior healing capacity of shTWIST1 hASCs-treated calvarial defects as compared to scramble, controls, and shTWIST21/TAZ hASCs-treated calvarial defects (Fig. 6E). Pentachrome staining of shTWIST1 hASCs-treated defects, in which bony regenerate appears yellow, revealed robust healing at week 10 after treatment, and the thickness of the regenerated bones in the defects was augmented. In contrast, scramble treatment showed less bone regeneration in the defect compared to shTWIST1 hASCs-treated defects (Fig. 6E). Reduced bone regeneration was observed in shTWIST1/TAZ-treated defects as well (Fig. 6E). Thus, shTWIST1hASCs had remarkable healing capacities, which were superior to scramble and shTWIST1/TAZ hASCs, confirming the greater osteogenic capacity observed previously in vitro.

Discussion

MSCs are currently being explored as vehicles for cell-based skeletal therapies [1, 2, 63, 64], therefore, investigating the mechanism(s) that fine-tune the balance between MSC, osteoblast, and adipocyte differentiation is likely to be of medical importance.

This study highlights the role of bHLH transcription factor TWIST1 in hASCs osteogenesis. TWIST1 can act as an inhibitor or inducers of osteogenesis depending on the tissue context. Earlier studies reported that Twist1 or -2 deficiency leads to premature osteoblast differentiation [38] suggesting an inhibitory effect on mesenchymal cells.

Our data show that TWIST1 silencing increased the osteogenic potential of hASCs in vitro and in vivo thus, indicating an inhibitory osteogenic activity of this transcriptional factor in hASCs. Of note, our analysis on ASCs derived from haploin-sufficient Twist1 mice also showed increased osteogenic activity of these cells compared to wild type ASCs (data not shown).

In this study, we found that TWIST1 was endogenously expressed at high levels in undifferentiated hASCs. Analysis of the TWIST1 expression profile performed on hASCs undergoing to osteogenesis showed that levels of TWIST1 expression gradually decreased during osteogenic differentiation. This is consistent with previous findings indicating that during calvaria and suture development, as well endochondral fetal bone development, Twist1 was expressed by early mouse pre-osteoblasts, but as these cells differentiated the expression dramatically decreased [32, 38, 65, 66]. Thus, TWIST1 expression in hASCs differentiating toward the osteogenic lineage recapitulated the embryonic profile observed during bones development.

We also observed that TWIST1 expression inversely correlated with the osteogenic markers RUNX2 and BGLP. These results suggest that TWIST1 may maintain hASCs in an undifferentiated state therefore, preventing them from differentiating toward the osteoblast lineage. In the light of this possibility, and to verify whether endogenous TWIST1 affects the osteogenic differentiation potential of hASCs, we downregulated its expression using a shRNA experimental approach. Successful transduction of TWIST1 shRNA into hASCs was demonstrated by the significantly lower expression of TWIST1 mRNA and TWIST1 protein as well.

Upon TWIST1 silencing, hASCs acquired the ability to differentiate along the osteogenic lineage in a more robust fashion both, in vitro and in vivo, compared to scramble hASCs. shTWIST1 hASCs were found to express higher levels of RUNX2, at early time point of osteogenic differentiation, which would suggest that decreased TWIST1 triggers upregulation of this osteogenic marker. This finding is in agreement with a study performed in zebrafish, showing that inhibition of Twist1a and Twist1b by morpholino oligonucleotides increased the expression of Runx2, whereas overexpression of Twist1a and Twist1b decreased it [39].

Twist1 has been reported to inhibit Runx2 function during skeletogenesis, suggesting that it may exert its role through multiple mechanism(s). Bialek et al. have demonstrated nuclear direct interaction between a specific region of Twist1 protein named the Twist box, with the DNA-binding domain of the transcriptional factor Runx2 resulting in the inhibition of this key transcriptional factor regulating osteogenic gene expression [38].

At later time points of osteogenic differentiation, higher expression of BGLP, a target gene for RUNX2 activity, was detected in shTWIST1 hASCs. Moreover, significantly more ALPL enzymatic activity and mineralization of extracellular matrix were observed in shTWIST1 hASCs as compared to scramble hASCs. Our results support previous findings which demonstrated an increase in RUNX2, ALP, and BGLP gene expression in siRNA knockdown of Twist1 in C3H10T1/2 cells, a murine mesenchymal cells line, as well in human periodontal ligament cells [67, 68], and in Twist1 antisense knockdown osteosarcoma cell lines [69]. Moreover, counterpart experiments where overexpression of TWIST1 was associated with a decrease in the gene expression of osteoblast markers BMP-2, ALPL, and BGLP [70] lend further support to our results.

Osteogenic inhibitory activity is a feature shared by several transcription factors, such as myeloid elf-1-like factor (MEF) [71], nuclear factor kappa B (NF-κB) [72], and others. These transcription factors can trigger inhibition of osteogenesis at different levels. For instance, MEF inhibits osteogenesis by directly interacting with Runx2 and suppressing its transcriptional activity in osteoblasts [71], while NF-κB exerts its inhibitory function on mesenchymal cells osteogenesis by promoting β-catenin degradation [72]. Therefore, in the context of cell-based bone regeneration, underpinning the molecular mechanism(s) which enhance the osteogenic potential of MSCs seizes a great opportunity for successful bone repair.

Development of bone and adipose tissue are linked processes arising from a common progenitor cell, but having an inverse relationship. Cellular differentiation of both tissues relies on growth factor cues. BMPs play critical roles in the commitment of mesenchymal cells into osteoblast and chondroblast lineages [73, 74], and BMP-2 has been found to significantly enhance osteogenic differentiation of hASCs [75, 76]. Other studies have indicated that FGF signaling also plays an important role for maintenance of the osteogenic potential of ASCs [52, 77]. Our results showing enhanced activation of both, BMP and FGF signaling pathways in shTWIST1 hASCs, define part of the molecular mechanism(s) underlying increased osteogenic differentiation elicited by TWIST1 silencing in these cells. The increased levels of BMPRIB and BMP-2 ligand observed in shTWIST1 hASCs are likely to be responsible for the enhanced activation of BMP signaling in these cells. This is strongly supported by finding that noggin treatment suppressed the enhanced activation of BMP signaling and osteogenesis in shTWIST1 hASCs.

Our data support previous studies suggesting that Twist1 may likewise affect BMP signaling in the process of mesenchymal cell differentiation into osteoblasts. A study by Hayashi et al. showed that overexpression of Twist1 suppresses BMP-induced osteoblast differentiation by recruiting histone deacetylases to Smad4, a known common signaling component associated with BMP signaling pathway [78]. On the contrary, downregulation of endogenous Twist1 enhances BMP signaling, thus indicating that Twist1 can act as an inhibitor of BMP signaling [78].

Furthermore, Twist1 has been suggested to be an upstream regulator of FGF receptors in cranio-facial and limb development [43] and to affect their transcription [79]. A study indicated that the positive effect of Twist1 silencing on osteogenic differentiation of mesenchymal cells is mediated in part via FGF receptor-2 (FGFR2) signaling [67]. In our study, we observed higher levels of FGR2 gene expression and protein as compared to scramble hASCs, which would suggest its role in the activation of ERK1/2-FGF signaling. Furthermore, pharmacological treatment with an inhibitor of ERK1/2 signaling pathway suppressed the enhanced osteogenic ability of shTWIST1 hASCs.

Previous studies have provided only a partial explanation on the positive effect of Twist1 silencing on osteoblastogenesis. Thus, a remaining question was to determine the downstream signaling mechanisms that mediate the positive effect of Twist1 silencing on osteogenic differentiation in mesenchymal cells. With our study, we progress a step forward unveiling further molecular component(s) involved in the positive effect elicited by TWIST1 silencing on hASCs osteogenesis. Our results identify the mesenchymal transcriptional coactivator TAZ as mediator of TWIST1 silencing effect on hASCs. This is strongly supported by results observed both in vitro and in vivo upon cosilencing TWIST1 and TAZ in hASCs.

Terminal differentiation of MSCs is largely controlled by the selective activation of specific programs of gene expression [80, 81]. These genetic programs are mutually exclusive and are triggered by key transcription factors, among them TAZ. This transcriptional coactivator with PDZ-binding motif plays a vital role in the osteogenic differentiation of MSCs and functions as a molecular rheostat, promoting MSCs differentiation into osteoblastic lineages by coactivating Runx2/ Cbfa1 and blocking MSCs differentiation into adipocyte lineages by repressing peroxisome proliferator-activated receptor γ-dependent gene transcription [59, 60]. Moreover, there is evidence showing that osteoblast-targeted overexpression of TAZ increases bone mass in vivo [82].

It has been shown that in BMP-2-induced osteoblastic differentiation of MSCs, TAZ level increased by several fold within a few days [60]. Furthermore, a recent study reported that FGF2 ligand stimulated MSCs osteogenic differentiation through ERK-induced TAZ expression [61]. Increased TAZ expression induced by IGF1 is also largely mediated by MEK-ERK pathway [83]. In shTWIST1 hASCs TAZ expression was dramatically reduced when cells were treated with inhibitors of either BMP or ERK1/2-FGF signaling pathways. These results would suggest that upregulation of TAZ observed in shTWIST1 hASCs is induced by these two converging signaling.

Because in vitro differentiation analysis often tells little about the ability of MSCs to differentiate in vivo, we reasoned to test the osteogenic potential of shTWIST1 hASCs in vivo as well. Using an established in vivo model of calvarial bone repair [13, 14], we demonstrated that shTWIST1 hASCs seeded on scaffold and implanted into a critical calvarial defect created in mice, significantly increased de novo bone regeneration when compared with defects treated with scramble hASCs or empty scaffold. Thus, the enhanced osteo-genic differentiation ability of shTWIST1 hASCs was not restricted in vitro, indeed occurred in vivo as well. Of note, all previous studies reported the effect of Twist1 silencing on osteogenic differentiation of cells tested in vitro only [67, 68]. Our results clearly demonstrate that shTWIST1 hASCs retained their greater osteogenic potential when tested in vivo to repair bony tissue. These findings have potential translational applications, it is thus foreseeable that cell-based strategies for bone tissue engineering may one day incorporate targeted silencing of TWIST1 to enhance skeletal regeneration.

Our functional studies demonstrate that TWIST1 acts as endogenous suppressor of MSC differentiation into osteoblasts. Furthermore, the observation that shTWIST1 hASCs exhibited a reduced capacity for adipogenesis compared to their corresponding scramble control strongly suggests that TWIST1 silencing modulates hASCs to bone fate switching rather than adiposity (Supporting Information Fig. S3). These findings support previous study showing that enforced expression of TWIST1 significantly increased adipogenic capacity in human MSCs [70]. Thus, TWIST1 may function as a temporal switch governing the fate of MSCs. Therefore, it is tempting to speculate that TWIST1 is another “molecular rheostat” counteracting TAZ. Level of TWIST1 can influence osteogenic gene expression and may act as a master switch in initiating bone cell differentiation by regulating the osteogenic cell lineage. This is strongly suggested by the downregulation of TWIST1 expression observed in hASCs undergoing osteogenic differentiation.

Our study provides more insights into the molecular mechanism(s) governing the enhanced osteogenic ability of shTWIST1 hASCs by identifying TAZ as a downstream effectors mediating their increased osteogenic potential. This is strongly supported by findings that TAZ expression was upregulated in TWIST1 knockdown hASCs and that either inhibition of BMP or ERK/FGF signaling abolished its upregulation, therefore leading to dramatic decrease in osteogenic differentiation of shTWIST1 hASCs. Overall, this study advances our understanding of the molecular mechanism(s) that drive enhanced osteo-genesis in hASCs, upon knocking-down TWIST1, and translates our basic findings to potential clinic applications.

Conclusions

Collectively, these findings indicate that TWIST1 silencing enhanced the osteogenic potential of hASCs both in vitro and in vivo, and reveal that this effect results from molecular mechanisms involving activation of endogenous BMP and ERK/FGF signaling pathways leading to upregulation of the molecular rheostat TAZ and enhanced osteogenic differentiation (Fig. 7). Moreover, our results place TAZ as a downstream component mediating the positive osteogenic effect elicited by knocking down TWIST1 in hASCs. This provides new insights into the molecular mechanism(s) through which TWIST1 controls osteogenic differentiation in mesenchymal cells and how to optimize the use of hASCs for skeletal regenerative medicine.

Figure 7.

Schematic proposed model depicting how TWIST1 knockdown enhances osteogenesis. TWIST1 silencing obtained by shTWIST1 technique acts in part, by inducing activation of the two endogenous osteogenic signaling BMP and ERK/FGF in hASCs, leading in turn to upregulation of the mesenchymal transcriptional coactivator factor TAZ, and therefore, to enhanced osteogenic differentiation. Abbreviation: hASCs, human adipose tissue-derived stem cells.