A Novel Strategy for Enrichment and Isolation of Osteoprogenitor Cells from Induced Pluripotent Stem Cells Based on Surface Marker Combination

In this study, we developed a new method to stimulate osteogenic differentiation in tissue-nonspecific alkaline phosphatase (TNAP)-positive cells liberated from human induced pluripotent stem cells (hiPSCs)-derived embryoid bodies (EBs) with 14 days long TGF-β/IGF-1/FGF-2 treatment. TNAP is a marker protein of osteolineage cells. We analyzed and isolated TNAP-positive and E-cadherin-negative nonepithelial cells by fluorescence-activated cell sorting. Treating the cells with a combination of transforming growth factor (TGF)-β, insulin-like growth factor (IGF)-1, and fibroblast growth factor (FGF)-2 for 14 days greatly enhanced TNAP expression and maximized expression frequency up to 77.3%. The isolated cells expressed high levels of osterix, which is an exclusive osteogenic marker. Culturing these TNAP-positive cells in osteoblast differentiation medium (OBM) led to the expression of runt-related transcription factor 2, type I collagen, bone sialoprotein, and osteocalcin (OCN). These cells responded to treatment with activated vitamin D3 by upregulating OCN. Furthermore, in OBM they were capable of generating many mineralized nodules with strong expression of receptor activator of NF-kappaB ligand and sclerostin (SOST). Real-time RT-PCR showed a significant increase in the expression of osteocyte marker genes, including SOST, neuropeptide Y, and reelin. Scanning electron microscopy showed dendritic morphology. Examination of semi-thin toluidine blue-stained sections showed many interconnected dendrites. Thus, TNAP-positive cells cultured in OBM may eventually become terminally differentiated osteocyte-like cells. In conclusion, treating hiPSCs-derived cells with a combination of TGF-β, IGF-1, and FGF-2 generated TNAP-positive cells at high frequency. These TNAP-positive cells had a high osteogenic potential and could terminally differentiate into osteocyte-like cells. The method described here may reveal new pathways of osteogenesis and provide a novel tool for regenerative medicine and drug development.


Introduction
The development of new treatment strategies for osteoporosis and other skeletal tissue diseases has become increasingly important considering the growing population of elderly people. Regenerative medicine and the development of new moleculartargeted agents are aimed at providing novel tools to address these clinical demands. Induced pluripotent stem cells (iPSCs) have attracted the attention of basic and clinical researchers since their establishment because they have the potential to provide useful tools for regenerative medicine and drug development. Before the development of iPSCs, human mesenchymal stem cells (hMSCs) were promising candidates for bone engineering and regeneration, and many successful studies with these cells have been reported. However, hMSCs have several limitations. hMSCs obtained from elderly people are generally low in number, grow slowly, and show diverse differentiation potentials. Utilization of hMSCs for drug development is difficult because of their limited proliferative ability and the poor reproducibility of the method. These problems could be resolved using human iPSCs (hiPSCs). However, the osteogenic differentiation of hiPSCs presents numerous problems, including time-consuming methods, poor reproducibility, and low efficiency. The designed differentiation of hiPSCs into osteolineage cells remains difficult and impedes progress. Several reports have described the directed differentiation of iPSCs or embryonic stem cells (ESCs) into multipotent progenitors or osteoprogenitors [1,2]. MSCs or MSC-like cells can be obtained from human ESCs by methods, such as fluorescence-activated cell sorting (FACS) after embryoid body (EB) formation. These protocols require prolonged serial passages or multiple cell sorting steps and are laborintensive, time-consuming, and generally inefficient [3].
Other skeletal tissues, such as muscles, can also be successfully generated from hiPSCs [4]. Goudenege et al. reported that hiPSCderived MSCs can be efficiently induced to undergo myogenic differentiation with MYOD1 overexpression [5]. However, these protocols have low reproducibility, probably because of the heterogeneous populations of MSCs that are derived from hiPSCs.
The other potential approach for generating skeletal tissues is to isolate paraxial mesodermal progenitors, which may differentiate into myogenic, osteogenic, and chondrogenic tissues [6]. Plateletderived growth factor receptor-a-positive and KDR-negative cells are immature, and thereby can differentiate into multiple types of tissues. Platelet-derived growth factor receptor-a-positive cells are partially differentiated and can be directed to differentiate into osteolineage cells. Tanaka et al. reported that MYOD1 overexpression in immature hiPSCs stimulates them to become mature myocytes with very high efficiency and reproducibility [7]. Their method provides relatively uniform undifferentiated cells, which may preclude variation in their differentiation frequency. Their results suggested that obtaining relatively uniform types of cells as early as possible may be very important.
We developed a new strategy to purify osteoprogenitors from EB-derived cells by isolating tissue-nonspecific alkaline phosphatase (TNAP)-positive cells using FACS. We found that cells separated from EBs did not express TNAP immediately after single-cell separation. They did not express E-cadherin but expressed relatively high levels of CD90, indicating that they were not progenitors of liver or bile duct epithelial cells. Treating the cells with a combination of transforming growth factor (TGF)b, insulin-like growth factor (IGF)-1, and fibroblast growth factor (FGF)-2 greatly enhanced TNAP expression. Furthermore, the cells began to express high levels of osterix (OSX), which is an exclusive osteogenic marker. The cells initially expressed low levels of runt-related transcription factor 2 (RUNX2), and continuous culture induced high levels of RUNX2, bone sialoprotein (BSP), type I collagen (COL1A1), and eventually, osteocalcin (OCN). To the best of our knowledge, these are the first observations of osteoprogenitors expressing high levels of TNAP and OSX but low levels of RUNX2 and collagen1a. In general, MSCs in vivo first express RUNX2, which promotes the expression of several early osteogenic marker proteins [8]. These RUNX2-expressing precursors then express OSX and induce differentiation of these cells into mature and functional osteoblasts. Therefore, OSX is a target molecule of RUNX2. However, in our experiment, OSX may have functioned as an initial transcription factor to initiate osteogenesis. We also found that these cells could form multiple mineralized nodules with multidendritic cells that express high levels of receptor activator of NF-kappaB ligand (RANKL), suggesting they can terminally differentiate into osteocyte-like cells. These cells are easily obtained from iPSCs and are capable of differentiating into osteocyte-like cells; they responded to treatment with activated vitamin D3 by upregulating OCN, providing a new clue in the investigation of osteocytes.

EB formation and in vitro differentiation
The differentiation method is shown in Figure 1. hiPSC colonies were dissociated with a cell scraper and transferred to lowattachment Petri dishes to generate EBs. EBs were maintained in suspension culture in human ES medium without FGF-2 for 6 days. EBs were then cultured in human ES medium with 2 mM thiazovivin without FGF-2 for 1 h at 37uC. After preincubation with 2 mM thiazovivin, EBs were collected and dissociated in 0.5 mg/ml collagenase type IV (Wako Pure Chemical Industries Ltd.) for 20 min at 37uC, followed by incubation in 0.05% trypsin-EDTA (Invitrogen) for 5 min at 37uC. The trypsinized EBs were seeded onto cell culture dishes at a density of 1.8610 4 cells/cm 2 and cultured in osteoblast differentiation medium (OBM), which consisted of a-MEM (Invitrogen) supplemented with 10% FBS, 50 mg/ml L-ascorbic acid (Wako Pure Chemical Industries Ltd.), 10 mM b-glycerophosphate (Wako Pure Chemical Industries Ltd.), and 10 nM dexamethasone (Wako Pure Chemical Industries Ltd.). Various combinations of cytokines [25 ng/ml FGF-2, 1 ng/ml TGF-b1 (Wako Pure Chemical Industries Ltd.), 100 ng/ml IGF-1 (Wako Pure Chemical Industries Ltd.), and 50 ng/ml bone morphogenetic protein (BMP)-2/-7 (R&D Systems, Minneapolis, MN, USA)] were added on the following day (day 0) and cultured for 14 days. OBM containing fresh cytokines was resupplied every 3 days. Human periodontal ligament cells (HPDLCs; Lonza, Basal, Switzerland) were cultured in OBM for 5 days. We previously reported that HPDLCs could differentiate into osteogenic cells by culturing in OBM [10].

Alkaline phosphatase (ALP) activity staining
Two weeks after stimulation, the cells were washed two times with phosphate-buffered saline (PBS), fixed in 4% paraformaldehyde for 5 min at room temperature, and washed three times with water. For staining, an ALP substrate solution (Roche Diagnostics, Basel, Switzerland) was added to the fixed cells for 60 min at room temperature. After staining, the cells were washed three times with distilled water, and the images were analyzed.
Antibodies, cell staining, flow cytometric analysis, and cell sorting After 2 weeks of osteogenic differentiation, cells from hiPSCderived EBs that had differentiated in culture in OBM were trypsinized with 0.05% trypsin-EDTA for 10 min at 37uC. The trypsinized cells were stained with anti-human ALP phycoerythrin-conjugated antibody (R&D Systems) for 45 min on ice in the dark. After staining, the cells were washed three times with PBS, suspended in PBS containing 0.5% FBS, passed through a 40mm mesh filter, and maintained at 4uC until flow cytometric analysis and cell sorting. Dead cells were excluded from flow cytometric analysis on the basis of propidium iodide staining (2 mg/ml) and forward scatter. We used a FACSAria (Becton-Dickinson, San Jose, CA, USA) which is a high speed cell sorter for measuring and sorting fluorescently labeled cells. Because a FACSAria is compatible with analyzing and sorting cells at the same time, we used a FACSAria to sort TNAP-positive cells. These TNAPpositive cells were found in cells cultured for 14 days in OBM supplemented with TGF-b, IGF-1, and FGF-2. After this cultivation in OBM, we sorted TNAP-positive cells by FACS.

Histochemistry for osteogenesis
Alizarin Red staining was performed as described previously [10]. In brief, the cultured cells were fixed with 4% paraformaldehyde in PBS for 5 min at room temperature, washed two times in PBS, incubated in Alizarin Red S solution for 5 min at room temperature, and washed five times in PBS at room temperature. Images were captured using a phase-contrast microscope.

Immunohistochemistry
The cells were fixed with 4% paraformaldehyde in PBS for 1 h. After washing, nonspecific binding of antibodies was blocked with 5% BSA in Tris-buffered saline with 0.05% Tween 20 (TBST, pH 7.6) at room temperature for 1 h. The cells were incubated with the primary antibody in TBST containing 5% BSA for overnight at 4uC (1:100 for goat anti-RANKL and 1:50 for rabbit anti-SOST; both from Santa Cruz Biotechnology, Dallas, TX, USA). The secondary antibodies were fluorescein isothiocyanateconjugated anti-goat IgG (1:200; Santa Cruz Biotechnology) and fluorescein isothiocyanate-conjugated anti-rabbit IgG (1:200; Zymed Laboratories, South San Francisco, CA, USA). The secondary antibodies were diluted in TBST and the cells were incubated for 1 h in the dark. The cells were finally stained with DAPI for nuclear staining.

Electron microscopy
For scanning electron microscopy (SEM), after FACS sorting, both TNAP-positive and -negative cells were cultured in OBM for 120 days. The cells were fixed in 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M cacodylate buffer, pH 7.4, for overnight at 4uC. After washing with 0.1 M cacodylate buffer, the samples were dehydrated in a graded series of ethanol and immersed in t-butyl alcohol for 30 min at 4uC. Next, the samples were freeze-dried with an ID-2 freeze dryer (Eiko, Tokyo, Japan) and sputter-coated with a cool sputter coater (SC500A, VG Microtech, East Sussex, UK). Finally, the samples were examined under a scanning electron microscope (SU-6600, Hitachi Co., Tokyo, Japan). For transmission electron microscopy (TEM), after FACS sorting, TNAP-positive cells were cultured in OBM for 120 days. The cells were fixed in 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M cacodylate buffer, pH 7.4, for overnight at 4uC. After washing with 0.1 M cacodylate buffer, the samples were postfixed with 1% osmium tetroxide in cacodylate buffer for 1 h at room temperature. The samples were then dehydrated and embedded in epoxy resin (Epon 812, Taab, Aldermaston, UK) for thin sectioning. Semi-thin sections were stained with toluidine blue for light microscopy. Thin sections were stained with uranyl acetate and lead citrate and examined by TEM (H-7650, Hitachi Co.).

Statistical analysis
Data are expressed as mean 6 S.D. and were analyzed using ANOVA, the Bonferroni test, or Student's t-test. All data are representative of at least three independent experiments. Statistical significance was defined as p,0.05.

TNAP-positive cells derived from human iPSCs
We induced EBs from hiPSCs and then the trypsinized EBs. TNAP-positive cells were not observed immediately after trypsinization.
The trypsinized EBs were cultured in OBM with various cytokines. We investigated which combination of cytokines was most effective in inducing TNAP-positive cells. The number of TNAP-positive cells was markedly increased by treatment with a combination of FGF-2, TGF-b, and IGF-1, and these cells could be isolated using FACSAria (Fig. 2a, b). ALP isoenzymes are categorized into four classes: germ cell specific, placenta specific, intestine specific, and tissue nonspecific. Parental hiPSCs and ALP-positive cells derived from hiPSCs by EB formation expressed the TNAP isoenzyme, whereas ALP-negative cells did not express any ALP isoenzyme (Fig. 2c). Subsequently, TNAP-positive cells were assessed for the expression of the MSC marker CD90 and the epithelial cell marker E-cadherin. Almost all cells were also CD90-positive and Ecadherin-negative (Fig. 2d). As shown in Figure 2e, TNAP-positive cells were spindle shaped, like MSCs, and TNAP-negative cells were cuboidal, like epithelial cells. These data suggested that TNAPpositive cells derived from hiPSCs were not epithelial cells.

TNAP-positive cells expressed various osteoblast marker genes
We determined the expression of ES cell markers and osteoblast-specific markers to confirm osteoblast differentiation. The expression of ES cell markers OCT3/4, SOX2, NANOG, REX1, ESG1 (DPPA5), and TERT was markedly reduced in both TNAP-positive and -negative cells compared with hiPSCs (Fig. 3a). In TNAP-positive cells, we found marked expression of not only TNAP but also OSX (Fig. 3b). Although the expression of RUNX2 and COL1A1 was relatively low, the expression of the marker for committed osteoblasts OSX was markedly increased, suggesting that these cells were osteolineage cells. Compared with TNAPnegative cells, the expression of the osteoblast-specific markers TNAP, OSX, BSP, and OCN was markedly upregulated in TNAPpositive cells cultured in OBM for 40 days (Fig. 3c). The expression of osteoblast markers RUNX2, COL1A1, and BSP was markedly increased in TNAP-positive cells when these cells were cultured for 40 days. In addition, OCN expression showed a  significant increase and TNAP expression showed a decreasing trend with vitamin D3 treatment (Fig. 3d). These observations indicated that these cells could respond to osteogenic reagents and could differentiate into cells in the late phase of osteogenesis.

TNAP-positive cells expressed characteristics of osteocyte-like cells
After culture in OBM for 40 days, TNAP-positive cells differentiated into osteocyte-like cells containing an abundant calcium matrix, as revealed by Alizarin Red-positive staining in the well (Fig. 4a). In contrast, TNAP-negative cells exhibited no potential to form calcium-positive osteocytes, as indicated by the absence of Alizarin Red staining (Fig. 4a). Many mineralized nodule-like structures were observed in cultures of TNAP-positive cells but not in those of TNAP-negative cells. In addition, TNAPpositive cells expressed RANKL in the areas of mineralized nodule-like structures. The expression of SOST was observed only in TNAP-positive cells but not in TNAP-negative cells (Fig. 4b).
The expression of osteocyte markers SOST, RELN, and NPY was significantly increased in TNAP-positive cells (Fig. 4c). The expression of these osteocyte markers increased concentrationdependently with vitamin D3 administration only after 6 days in culture (Fig. 4d). As shown in Figure 4e, expression of the osteocyte marker genes DMP1, FGF23, and MEPE was detected in TNAP-positive cells.

Morphology of osteocyte-like cells as observed by electron microscopy
After culture in OBM for 120 days, many cytoplasmic processes were observed in TNAP-positive cells (Fig. 5a). In contrast, TNAPnegative cells were round and had no cytoplasmic processes (Fig. 5a). Cell-cell contact with a cytoplasmic process was observed in TNAP-positive cells (Fig. 5b, c, arrowheads). Discussion iPSCs are powerful tools in many fields of basic scientific research. Several reports have shown that osteogenic cells can be generated from iPSCs [3,6,7,[11][12][13][14][15]. The reported methods for the generation of osteogenic cells are time-consuming and laborintensive and include repeated passages to select fast-growing adhesive cells [2]. The phenotypic characteristics of these cells are similar to those of mesenchymal cells. Bilousova et al. [14] reported that retinoic acid treatment of murine iPSCs cultured in OBM for several weeks resulted in cells that were positive for osteogenic markers and Alizarin Red staining. This so-called outgrowth method essentially requires no supplements other than OBM. However, human iPSCs are not as simple to differentiate as murine iPSCs. Multistep, labor-intensive processes are often necessary. Mahmood et al. [16] reported that iPSCs that were cultured in low-adhesive plastic Petri dishes with the TGF-b inhibitor SB-431542 for 10 days formed EBs and adhered to the cell culture dishes. These cells could be passaged 4-11 times. The cells were then transferred into OBM and cultured for an additional 20 days, eventually forming osteoblasts. Villa-Diaz et al. [3] used synthetic polymer-coated dishes to generate MSCs. It is possible that these MSCs derived from iPSCs were a mixed population of cells, although the protocol usually requires a long period of time. Thus, methods that are simpler and less timeconsuming are desired.
The most essential proteins for mineralization by osteolineage cells are COL1A1 and ALP [17]. Humans have four ALP genes encoding intestinal, placental, placenta-like, and liver/bone/ kidney (i.e., TNAP) gene products. TNAP is localized on the outside of the plasma membrane of cells and in the membrane of matrix vesicles and is attached to the membrane by a glycophosphatidylinositol anchor. TNAP in osteolineage cells is expressed relatively early during differentiation and is abundantly expressed on the membrane surface. Apart from osteolineage cells, epithelial cells also express TNAP. Therefore, accumulation of relatively pure osteolineage cells among TNAP-positive cells will probably require elimination of epithelial cells. Some cells among EBs derived from hiPSCs expressed ALP (Fig. 2a, b). RT-PCR clearly showed that these cells expressed TNAP (Fig. 2c). Morphologically, TNAP-positive cells were fibroblastic and spindle shaped rather than cuboidal or epithelial shaped (Fig. 2e). We next examined whether these cells expressed E-cadherin. Most cells were E-cadherin-negative and CD90-positive, indicating that TNAP-positive cells are probably not epithelial cells. We further investigated methods of maximizing TNAP expression. Previous reports indicated that activins, retinoic acids, and BMPs are capable of inducing osteolineage cells from ESCs or iPSCs [6,13,14]. We attempted several combinations of cytokines and found that activins, retinoic acids, and BMPs did not effectively induce TNAP expression (data not shown). However, exposure to TGF-b, IGF-1, and FGF-2 in OBM had the most potent TNAPinducing effects (Fig. 2a). Although the mechanisms by which ALP expression is regulated are complex, the BMP/Runx2 and Osx systems are believed to be the principal regulatory pathways controlling osteoblast differentiation and TNAP expression [8]. In brief, the BMP/Smad pathway targets activation of Runx2, which in turn activates Osx expression. Because epigenetic conditions during embryonic development are quite different from those during ESC/iPSC differentiation, the transcription factors required for TNAP expression may be different. In embryos, Runx2 is required for the differentiation of prechondrogenic mesenchymal cells into osteoblasts, whereas Osx is believed to induce subsequent maturation of osteoblasts and inhibit chondrogenic differentiation. In Osx-null embryos, cartilage forms normally but the embryos completely lack bone [17][18][19]. OSX, which is specifically and exclusively expressed in all osteoblasts, showed markedly high expression in TNAP-positive cells, although TNAPpositive and -negative cells expressed almost similar levels of RUNX2 (Fig. 3b). These findings indicated that iPSCs may not require the prechondrogenic process and may induce Osx without a Runx2 surge. Several pathways have been reported to increase Osx expression. Mitogen-activated protein kinases, particularly p38, Erk1/2, and protein kinase D, activate Osx expression accompanied by TNAP activation. Ascorbate-dependent prolyl hydroxylase domain protein induces Osx expression. Endoplasmic reticulum stress also increases Osx induction [20]. These cascades may play an important role in the Osx surge and the increase in TNAP in iPSCs. We found that continuous culture of these TNAP-positive cells in OBM eventually led to increased expression of RUNX2, TNAP, COL1A1, and OSX as well as other osteogenic markers, such as BSP and OCN. These results indicated that TNAP-positive cells derived from hiPSCs are OSX-positive osteoprogenitors, not chondrogenic cells. Furthermore, TNAPpositive cells are capable not only of differentiating into osteogenic cells but also of responding to active vitamin D treatment. Vitamin D treatment effectively upregulated OCN and downregulated TNAP, indicating that these cells could differentiate into cells in the late phase of osteogenesis and may be able to differentiate into terminally differentiated osteocytes.
Previous reports showed that murine and human ESCs cultured in OBM formed a number of bone/mineralized nodules with intense mineralization [21,22]. A bone nodule is a group of cells with three-dimensional multistratified structures. We found that TNAP-positive cells derived from hiPSCs formed several bone nodules that contained intensely stained anti-RANKL-immuno-positive cells (Fig. 4b). We also observed anti-SOST positivity in these areas (Fig. 4b). qRT-PCR and RT-PCR clearly showed a significant increase in the expression of osteocyte marker genes, including SOST, NPY, RELN, DMP1, FGF23, and MEPE (Fig. 4c,  d, and e). SEM showed that TNAP-negative cells had a cuboidal morphology without dendritic structures, whereas TNAP-positive cells were flattened with multiple dendritic morphologies after cultivation in OBM (Fig. 5a). Toluidine blue-stained semi-thin sections clearly showed that these dendrites were connected to each other. The osteocyte-like cell line MLO-Y4 shows similar morphology [23]. Thus, these cells were osteocyte-like cells.
In the present study, formation of bone nodules was observed much more frequently in iPSCs than in MSCs, similar to the findings of a previous study [24]. Because ESCs and iPSCs have a higher proliferative potential, they may form multistratified structures more readily and thus provide cells with a threedimensional microenvironment that promotes terminal differentiation into osteocyte-like cells. Although our cells were positive for most osteocyte marker proteins, the expression of PHEX was not increased. This difference may be related to the in vivo microenvironment.
In conclusion, treating trypsinized single cells with a combination of TGF-b, IGF-1, and FGF-2 generated TNAP-positive cells at a high frequency. These TNAP-positive cells had high osteogenic potential and could terminally differentiate into osteocyte-like cells. They responded to osteogenic reagents and may be a useful tool for drug evaluation.