Dual usage of a stage-specific fluorescent reporter system based on a helper-dependent adenoviral vector to visualize osteogenic differentiation

We developed a reporter system that can be used in a dual manner in visualizing mature osteoblast formation. The system is based on a helper-dependent adenoviral vector (HDAdV), in which a fluorescent protein, Venus, is expressed under the control of the 19-kb human osteocalcin (OC) genomic locus. By infecting human and murine primary osteoblast (POB) cultures with this reporter vector, the cells forming bone-like nodules were specifically visualized by the reporter. In addition, the same vector was utilized to efficiently knock-in the reporter into the endogenous OC gene of human induced pluripotent stem cells (iPSCs), by homologous recombination. Neural crest-like cells (NCLCs) derived from the knock-in reporter iPSCs were differentiated into osteoblasts forming bone-like nodules and could be visualized by the expression of the fluorescent reporter. Living mature osteoblasts were then isolated from the murine mixed POB culture by fluorescence-activated cell sorting (FACS), and their mRNA expression profile was analyzed. Our study presents unique utility of reporter HDAdVs in stem cell biology and related applications.


Results
Propagation and titration of HDAdVs. The structures of HDAd-hOC-Venus and HDAd-CAG-Venus are shown in Fig. 1, Supplementary Fig. S1. HDAd-hOC-Venus encodes the Venus fluorescent reporter gene under the control of human OC locus while HDAd-CAG-Venus has a constitutively active strong promoter to drive Venus in almost every cell type. HDAd-hOC-Venus and HDAd-CAG-Venus vector was propagated for 4 rounds in 116 cells; after purification, the physical titer was 9.8 × 10 10 vector particle (vp)/ml and 8.7 × 10 10 vp/ ml, respectively (determined by quantitative Southern hybridization). The infectious titer of HDAd-hOC-Venus and HDAd-CAG-Venus vector was 3.7 × 10 10 β-gal-transducing unit (btu)/ml and 3.3 × 10 10 GFP-transducing unit (gtu)/ml, respectively. The multiplicities of infection (MOIs) of HDAd-hOC-Venus for various cell types were optimized based on the transduction efficiency of HDAd-CAG-Venus. MG-63 or HeLa cells were infected with the HDAdVs at an MOI of 2000 or 100 vp/cell, respectively. At this MOI, 99% of the cells were transduced by the HDAd-CAG-Venus vector ( Supplementary Fig. S2). Human induced osteoblasts or mouse POBs, on which bone-like nodules were formed, were infected with the HDAdVs at an MOI of 1,000. With this MOI, 60% of the cells (including both osteoblasts forming bone-like nodules and the, surrounding cells) were transduced with the HDAd-CAG-Venus vector ( Supplementary Fig. S3). Then, the promoter-dependent expression from hOC-Venus and its dose-dependency on 1α,25(OH) 2 D 3 (active vitamin D3; VD3) were examined by FACS and a quantitative reverse transcription polymerase chain reaction (qRT-PCR) in MG-63 cells. The mean fluorescence intensity (MFI) of the total cells ( Supplementary Fig. S4a,d) increased after VD3 treatment in a dose-dependent manner, in parallel with the mRNA levels of the Venus gene itself (Supplementary Fig. S4b) and of the OC gene ( Supplementary Fig. S4c) in MG-63 cells. www.nature.com/scientificreports www.nature.com/scientificreports/ The Venus expression in HDAd-hOC-Venus-infected bone-like nodules induced from human osteoblast cultures. To confirm the utility of our HDAdV vector, HDAd-hOC-Venus was used to directly infect human osteoblast cultures induced from the iPSC line SeVdp (KOSM) #7 (hereafter TIG3/KOSM) with the HDAd-hOC-Venus vector at an MOI of 1,000. The specific expression of the hOC-Venus construct at bone-like nodules was observed ( Supplementary Fig. S3), which was confirmed by alizarin red staining ( Supplementary  Fig. S7).

The Venus gene expression at bone-like nodules induced from OC-Venus knock-in hiPSC lines.
To compare the specificities of Venus expression from the HDAd-hOC-Venus vector and the endogenous OC expression, OC-Venus KI human iPSC lines were established by homologous recombination after the infection of the wild-type human iPSC line with the same HDAd-hOC-Venus vector (Fig. 2). After positive selection with G418 and negative selection with ganciclovir (GANC), 6 of 20 of G418/GANC double-resistant colonies were confirmed to be OC-Venus KI clones by a genomic PCR. The results from four representative KI clones are shown (Fig. 2c, Supplementary Fig. S5a). One of the OC-Venus knock-in human iPSC lines, KI #37 (OCVneo37), was transfected with pOG44, an Flp recombinase expression plasmid, to excise the neomycin-resistant gene cassette, because the presence of a drug-resistant gene may interfere with the regulation of the reporter gene expression. Six of 24 single colonies lost resistance to G418 after cloning. Five of them were confirmed to be neo  www.nature.com/scientificreports www.nature.com/scientificreports/ cassette-excision (EX) clones (Fig. 2d, Supplementary Fig. S5b). Two of the neo cassette-excision human iPSC lines, EX #10 (OCV37FF10) and #11 (OCV37FF11), were subjected to the following differentiation assays.
The unmodified control human iPSC line (TIG3/KOSM) and the EX lines (OCV37FF10 and OCV37FF11) were induced into neural crest-like cells (NCLCs) for 10 days, and then into mesenchymal stromal cells (MSCs) for 7 days. Bone-like nodules were formed after the induction of MSCs into the osteoblast lineage for 14 days (Fig. 3a). In cells without osteogenic induction, fluorescence was not detected by fluorescence microscopy (Fig. 3b,c) or FACS (Fig. 3d). The bone-like nodules of the control (CT) lines did not show Venus fluorescence (Fig. 3e), while those of EX lines did (#11; Fig. 3f, #10; Supplementary Fig. S6a). The fluorescence-emitting bone-like nodules were positively stained by alizarin red (Supplementary Fig. S7). The proportion of Venus-positive cells appeared to be no more than 1.5% (Fig. 3g, Supplementary Fig. S6b). Of note, the pattern of Venus expression was similar to that observed in human osteoblast cultures induced from the wild-type iPSC line and infected with HDAd-hOC-Venus, confirming the highly specific expression of HDAd-hOC-Venus.

The Venus expression in HDAd-hOC-Venus -infected bone-like nodules induced from mouse osteoblast cultures.
One of the advantages of the HDAdV is that it can also transduce cells from various species other than human. The mature osteoblast-specific expression of hOC-Venus was further examined in cultures of mouse POBs prepared from calvaria. In the cultures of POB, numerous bone-like nodules were formed by day 7. When these cultures were infected with the HDAd-hOC-Venus vector, Venus was specifically expressed in the cells forming bone-like nodules (Fig. 4a). Immunostaining using an anti-OC antibody also specifically labeled the bone-like nodules. The merged images of the HDAd-hOC-Venus expression and the anti-OC staining indicated that the expression of hOC-Venus was colocalized with that of endogenous OC. In contrast to hOC-Venus, CAG-Venus was expressed not only in the nodule-forming cells, but also in the surrounding cells, which were not stained with anti-OC antibody, confirming the specificity of the hOC-Venus expression (Fig. 4a).

A comparison of the Venus expression controlled by the 19-kb human OC locus and the 3.8-kb human OC promoter.
To compare the effects of the 19-kb OC locus in HDAd-hOC-Venus and the 3.8-kb OC promoter, used in other studies, in driving the reporter gene, HDAd-hOC-Venus, E1DAd-hOC3.8-Venus and E1DAd-CMV-GFP were used to infect mouse POB cultures at the same MOI of 1,000 (Fig. 5). E1DAd-hOC3.8-Venus and E1DAd-CMV-GFP are both E1-deleted AdVs expressing the reporter gene under the control of the 3.8-kb OC promoter and the CMV enhancer/promoter, respectively. The www.nature.com/scientificreports www.nature.com/scientificreports/ transduction efficiency of E1DAd-CMV-GFP was 63% (Fig. 5). While the expression of HDAd-hOC-Venus at the bone-like nodules of mouse POBs was strong and specific (Fig. 5), that of E1DAd-hOC3.8-Venus showed a much weaker signal (Fig. 5). The infectious titers of the two vectors are estimated to be similar or even higher for E1DAd-hOC3.8-Venus, because when they were used to infect HeLa cells at the same MOI of 100, the Venus-transduction efficiency of cells infected by E1DAd-hOC3.8-Venus was almost equal to or higher than that of cells infected by HDAd-hOC-Venus with 1, 10, 100 μM VD3 induction ( Supplementary Fig. S8). While the simple VD3-dependent expression was reproduced in HeLa cells by the minimal OC3.8 promoter with VDRE, the OC transcription might be regulated in a more complicated manner in the 19-kb OC locus.
Characterization of hOC-Venus-expressing primary mouse osteoblasts. We then characterized the expression profile of the osteoblast marker mRNAs in the bone nodule-forming osteoblasts. Because it is likely that those cells derived from mouse POBs are more genuine than those induced from hiPSCs, we chose the mouse POB culture system as a source of mRNA. The HDAd-hOC-Venus-infected mouse POB cells were sorted by FACS using the fluorescence intensity of Venus, since no Venus (+) cells were detected in the non-infected cultures (Fig. 4b). As the proportion of Venus (+) cells fluctuated from 5% (Fig. 4b) to 30% (Fig. 5) between the experiments, we repeated the experiments independently three times. The Venus (+) fraction in the HDAd-hOC-Venus infected cultures was 5.0% of the total cells (Fig. 4b), and the purity after FACS sorting was 87% ( Supplementary Fig. S9) in a representative experiment. The relative expression levels of mRNAs expressed in osteoblasts in the hOC-Venus (−) and hOC-Venus (+) cells were examined by a qRT-PCR. Regardless of some inconsistent results between experiments, reproducible patterns were observed (Fig. 6). Among the twelve genes examined, the expression levels of five endogenous genes (OC   Table S1. www.nature.com/scientificreports www.nature.com/scientificreports/ fold], Alp [5.6-9.7 fold] and Osx [2.2-8.1 fold]) in the Venus (+) cells were higher than those in the Venus (−) cells (Fig. 6). The expression of an osteogenic transcription factor, Runx2, was also slightly higher (1.1-2.4 fold) in hOC-Venus (+). In contrast, the expression levels of four genes of the secretory proteins (Col1a1, Col1a2, Spp1 and Sparc) and two genes related to bone resorption (Opg and Rankl) did not differ to a statistically significant extent between Venus (+) and Venus (−) cells.

Discussion
In the present study, we showed the dual usage of a novel adenoviral vector, HDAd-hOC-Venus, which allows for both the generation of knock-in reporter cell lines by homologous recombination in human iPSCs and postnatal visualization and separation of bone-forming mature osteoblasts by transient infection into mouse and human cells.
Our vector system has distinct advantages over other methods for tracing osteoblasts. First, our HDAdV-based system allows for the transient expression of Venus without chromosomal integration in the target cells. Thus, unlike transgenic animals or lentivirus-based systems, the vector DNA will be eventually lost from the infected cells after several cell divisions. This feature is potentially quite advantageous for the clinical application of the vector, such as for the enrichment of mature osteoblasts for transplantation. Second, adenovirus-mediated gene transfer is more efficient than other viral and non-viral methods in various cell types. In contrast to the GFP transgenic mice driven by the human OC promoter 4 , our HDAdV system is more versatile and can be used to visualize the OC-positive cells in unmanipulated animal species, including humans. Born et al. used nucleofection to deliver the 3.8-kb human OC promoter-GFP construct into human bone marrow-derived cells with a maximum transfection efficiency of 25% 5 ; thus, a double-gene construct of the OC promoter-GFP and CMV promoter-H2B-RFP had to be used to enrich transfected cells by sorting before their differentiation experiment 6 . However, the transduction efficiency of the double-gene construct was even lower (5.3-5.6%), probably because of the larger size of the construct, as discussed by the authors. Compared with nucleofection, the transduction efficiencies of adenoviral vectors of approximately 30 kb into mouse and human osteoblasts were as high as 60% at an MOI of 1,000, which was used in this study (Fig. 5, Supplementary Fig. S2). No reduction in the cell viability was observed around an MOI of 1,000, and we have observed no negative consequences due to adenoviral infection below an MOI of 10,000. Third, while antibody-based cell sorting is useful for proteins expressed on the cell surface, the reporter system with the promoter-specific expression of fluorescent protein allows us to sort target cells expressing proteins for which no suitable antibody is available as well as intracellular or secreted proteins. Although a previous study demonstrated that using OC antibodies for FACS is useful for detecting circulating osteoblasts 18 , the transient expression of Venus by OC-positive cells during induction into an osteogenic lineage allows us to evaluate the degree of differentiation via microscopy and conduct continuous experiments without fixing the cells for antibody staining. This is an important advantage, especially for bone research, which involves time-consuming and complicated induction processes. Fourth, the tissue-specific expression achieved using the HDAdV vector is superior to that achieved when using the first-generation adenoviral vector due to its high capacity to accept long regulatory elements of up to 20 kb and minimal viral enhancers, which will interfere with the expression of reporter genes 19 . In support of this, when POB forming bone-like nodules were infected with HDAd-hOC-Venus, Venus was specifically expressed in the bone-like nodules, which were stained with anti-OC antibody (Fig. 4a). In addition, the expression levels in the bone-like nodules that were achieved using the 19-kb human OC locus were higher than those achieved using the 3.8-kb OC promoter (Fig. 5). These results indicate that, to precisely reproduce tissue-specific expression of an endogenous gene, it is critical to use a large genomic locus to drive a reporter gene. HDAdVs have a cloning capacity of roughly 30 kb and are therefore an ideal vector for this purpose. Finally, using the same HDAdV, we were able to establish knock-in human iPS cell lines with which we can visualize the developmental expression patterns of the OC gene. Although the targeting efficiency when using infection with HDAdVs is reportedly lower than when using artificial nucleases 17 , the former approach is roughly 300-fold as efficient as that by traditional electroporation of naked plasmid DNA 13 . In addition, this vector system can obviously be used in combination with artificial nucleases to achieve even more efficient gene targeting 20 . Using high-capacity HDAdVs has additional advantages over traditional methods, including efficient transduction into a wide range of cell types, efficient knock-in of large DNA cassettes, simultaneous introduction of multiple modifications to a large DNA region, and no risk of off-target cleavage caused by artificial nucleases, as discussed previously 13,14,17 .
However, despite these advantages, one limitation associated with the HDAd-hOC-Venus vector might be that the infection efficiency could not reach 100% at an MOI of 1,000 in our experiment, even though this infection efficiency was superior to that of nucleofection. Thus, the hOC-Venus (−) fraction could contain endogenous OC-positive cells that were not infected. Infection at higher MOIs, at 10,000, resulted in significant cell death of POBs. To further improve our system, a higher infection efficiency without cytotoxicity -possibly through the use of adenoviral vectors from other serotypes or a method that excludes uninfected cells -is therefore required. Furthermore, propagating HDAdVs takes more time than traditional gene targeting methods using plasmid DNA and requires a biohazard level 2 facility.
In the present study, we successfully enriched OC-expressing mature osteoblasts from the mixed POB cultures using the HDAd-hOC-Venus virus. We found that OC-expressing mature osteoblasts co-expressed Bsp, Pth1r, Alp and Osx mRNAs at higher levels than in immature osteoblasts. The results are expected for Bsp because it has been reported to be highly expressed in mature mineralizing osteoblasts 21 , while Alp and Pth1r have been regarded as an early marker of osteogenesis. The expression of Alp and Pth1r might start at an early stage and increase during maturation, as reported in some in vitro analyses using the MC3T3-E1 cell line 22 . In addition, our results suggested that Osx is involved in the later stages of osteogenesis and that it regulates the expression of OC, Bsp, Pthr1 and Alp in mature osteoblast forming bone-like nodules in vitro. Further analyses of the binding sequences of Osx in the promoter/enhancer regions in those genes may therefore help us to understand the molecular mechanisms underlying such differentiation-dependent transcription of the genes related to bone formation.
www.nature.com/scientificreports www.nature.com/scientificreports/ In summary, we established a novel HDAdV-based system to detect living mature osteoblasts using the long regulatory sequences of the human OC gene. Our HDAdV-based hOC-Venus reporter will therefore be useful to visualize, isolate and characterize mature osteoblasts in various systems. Such HDAdV-based transient reporter systems would also be widely applicable to isolate other cell lineages or tissues from various species, including humans.  24 was cultured in MEM with 10% FCS supplemented with 1 mM sodium pyruvate (Sigma-Aldrich, St. Louis, MO) and 1x non-essential amino acids (Sigma-Aldrich). The human cervical carcinoma cell line, HeLa, and HEK293 cell line were cultured in DMEM (Nacalai tesque) with 10% FCS. The human iPSC line, TIG3/KOSM (formerly termed SeVdp (KOSM) #7) 25 was maintained as on-feeder 26 or feeder-free 27 culture, as described previously.

Construction and preparation of HDAdVs.
To generate the human OC-Venus HDAdV vector (HDAd-hOC-Venus), RP11-54H19, a BAC clone containing the human OC locus (BACPAC resources, Children's Hospital & Research Center at Oakland, Oakland, CA), was modified using the Red/ET recombination technique 28 . An FRT-PGK-EM7-neo-pA-FRT cassette was inserted into a single NotI (New England Biolabs, Ipswich, MA) site of pCS2-Venus (kindly provided by Dr. Atsushi Miyawaki) 9 . Next, the Venus-pA-FRT-PGK-EM7-neo-pA-FRT cassette was amplified by a PCR using primers with a 40-nt homology sequence to the target site (Table S1a upper) and inserted into exon 1 of the OC gene on the BAC (Fig. 1a-c). The ATG start codon of Venus was fused in-frame with the ATG of the OC gene. Subsequently, a total of 21.4-kb OC gene locus, including the marker cassette, was retrieved into a PCR-amplified pBR322 vector backbone with a second primer set (Table S1a lower). The entire cassette was excised by SalI (New England Biolabs) and inserted into an HDAdV plasmid, pAMHDAdGT8-4 14 . The resultant pAMHDAdGT-hOC-Venus plasmid was linearized by PmeI (New England Biolabs) (Fig. 1d) and packaged into virus particles (Fig. 1e) by transfection into 116 cells with the addition of AdNG163R-2 helper virus (kindly provided by Dr. Phillip Ng). The viral vector was propagated by serial passages in the 116 cell line with AdNG163R-2 and purified, as described previously 23,29,30 . The physical titer of the vector was determined as the copy number of viral genomic DNA by a quantitative Southern analysis 30 . The infectious titer was determined as β-gal-transducing units (btu) by X-gal staining on the 293LP cell line 29 . HDAd-CAG-Venus (formerly termed HDAdVenus-geo-TK 13 ) was also propagated and used as a constitutive Venus-expressing control ( Supplementary Fig. S1). The infectious titer of this vector was determined as GFP(Venus)-transducing units (gtu) measured on 293A cells by a FACS analysis. The MOI for each cell type was defined as the vector copy number to the cell number.

Generation of knock-in reporter hiPSC lines with HDAd-hOC-Venus vector.
A control hiPSC line, TIG3/KOSM, maintained on SNL feeder cells with hES medium was treated with CTK solution and suspended in hES medium as small clumps, as previously described 26 . They were then infected with HDAd-hOC-Venus at an MOI of 300 and plated onto SNL feeder cells. G418 selection (50 μg/ml; Nacalai tesque) was started 2 days after infection. After 3 weeks, the surviving colonies were transferred to 96-well plates and ganciclovir (GANC) selection (2 μM; Thermo Fisher Scientific, Waltham, MA) was started. G418/GANC double-resistant clones were characterized by a genomic PCR using the primers shown in Table S1b and Fig. 2 with PrimeSTAR GXL polymerase (Takara Bio, Japan), in accordance with the manufacturer's instructions. The resulting OC-Venus KI hiPSC lines were habituated to feeder-free culture using StemFit AK03 medium (Ajinomoto, Japan) and iMatrix-511 (Nippi, Japan) for two passages, as previously described 27 . They were then transfected with pOG44 (Thermo Fisher Scientific) using TransIT ® -LT1 Transfection Reagent (Takara Bio), according to the manufacturer's protocol, and plated onto iMatrix-511-coated 6-cm culture dishes. Single colonies were picked-up and the resistance to G418 was checked on duplicated plates. The clones that were sensitive to G418 were selected as candidate clones with neo cassette-excision (EX) by Flp recombinase, and the excision of the neo cassette was confirmed by a PCR using the primers shown in Table S1b and Fig. 2. Differentiation of human iPSCs into osteoblasts. The control hiPS cell line, TIG3/KOSM, and OC-Venus KI hiPS cell lines were cultured under Matrigel-coated feeder-free conditions with mTeSR-1 (BD Bioscience, San Jose, CA). NCLC induction was performed for 10 days, as previously described 31 . Induced NCLCs were maintained with medium containing DMEM (Sigma-Aldrich), 10% FBS (Thermo Fisher Scientific), 1% L-glutamine (Nacalai Tesque) for an additional 7 days. After additional culturing, the cells showed vigorous expansion and appeared to be MSCs. The MSCs were induced into osteoblasts with Osteogenic Differentiation Medium BulletKit ™ (Lonza, Switzerland) on a 6-well dish or an 8-well glass chamber slide II (AGC Techno Glass, Japan) for 14 days (Fig. 3a).
Primary osteoblast isolation and the nodule formation assay. Mouse POBs were isolated according to the published methods 2,32 . POBs were obtained from the calvariae of neonatal C57BL/6J Jc1 mice (Clea Japan, Japan). A mixture of 0.1% collagenase (FUJIFILM Wako, Japan) and 0.2% dispase was used to dissociate cells from the bone fragments. POBs were then seeded onto either a 6-well plate or an 8-well chamber slide II at a density of 2.0 × 10 6 or 3.2 × 10 5 cells/well, respectively. The cells were cultured in αMEM (Thermo Fisher Scientific) with 10% FCS supplemented with 50 U/ml streptomycin (Thermo Fisher Scientific) and 50 μg/ml penicillin (Thermo Fisher Scientific). The medium was changed every 48 h for 1~2 weeks until bone-like nodules were formed. www.nature.com/scientificreports www.nature.com/scientificreports/ Immunohistochemistry and microscopy. The human osteoblasts induced from the control iPSC line or mouse POBs infected with the HDAdVs were immunostained using a standard method with an anti-OC rabbit polyclonal antibody, LSL-LB-4005 (Cosmo Bio, Japan) as a primary antibody and Alexa Fluor 594-conjugated anti-rabbit IgG antibody (Thermo Fisher Scientific) as a secondary antibody. The cells were mounted with ProLong Gold antifade reagent with 4′,6-diamidino-2-phenylindole (DAPI) (Thermo Fisher Scientific), and were then examined under a fluorescence microscope, IX81 (Olympus, Japan), with a CCD camera, CoolSNAP HQ (Photometrics, UK) or ORCA-ER (Hamamatsu photonics, Japan).