Author contributions: Z.Y.Z.: design of the study, collection, analysis, and interpretation of data, manuscript writing; T.S.H.: conception and design of the study, provision of study material, analysis and interpretation of data, manuscript writing; M.C.: provision of study material, analysis and interpretation of data, manuscript writing; J.T.S.: provision of study material; N.M.F.: analysis of data, manuscript writing; M.C.: conception and design, financial support, administrative support, provision of study materials, analysis of data, manuscript writing, final approval of the manuscript; J.C.: conception and design, financial support, administrative support, provision of study materials, analysis of data, manuscript writing, final approval of the manuscript; M.A.C. and J.C. contributed equally to this work.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSExpress October 2, 2008.
Mesenchymal stem cells (MSCs) from human adult bone marrow (haMSCs) represent a promising source for bone tissue engineering. However, their low frequencies and limited proliferation restrict their clinical utility. Alternative postnatal, perinatal, and fetal sources of MSCs appear to have different osteogenic capacities, but have not been systematically compared with haMSCs. We investigated the proliferative and osteogenic potential of MSCs from human fetal bone marrow (hfMSCs), human umbilical cord (hUCMSCs), and human adult adipose tissue (hATMSCs), and haMSCs, both in monolayer cultures and after loading into three-dimensional polycaprolactone-tricalcium-phosphate scaffolds.Although all MSCs had comparable immunophenotypes, only hfMSCs and hUCMSCs were positive for the embryonic pluripotency markers Oct-4 and Nanog. hfMSCs expressed the lowest HLA-I level (55% versus 95%–99%) and the highest Stro-1 level (51% versus 10%–27%), and had the greatest colony-forming unit–fibroblast capacity (1.6×–2.0×; p < .01) and fastest doubling time (32 versus 54–111 hours; p < .01). hfMSCs had the greatest osteogenic capacity, as assessed by von-Kossa staining, alkaline phosphatase activity (5.1×–12.4×; p < .01), calcium deposition (1.6×–2.7× in monolayer and 1.6×–5.0× in scaffold culture; p < .01), calcium visualized on micro-computed tomography (3.9×17.6×; p < .01) and scanning electron microscopy, and osteogenic gene induction. Two months after implantation of cellular scaffolds in immunodeficient mice, hfMSCs resulted in the most robust mineralization (1.8×–13.3×; p < .01).The ontological and anatomical origins of MSCs have profound influences on the proliferative and osteogenic capacity of MSCs. hfMSCs had the most proliferative and osteogenic capacity of the MSC sources, as well as being the least immunogenic, suggesting they are superior candidates for bone tissue engineering. STEM CELLS2009;27:126–137
Mesenchymal stem cells (MSCs) are rare cells that can be readily isolated from bone marrow (BM) and expanded through multiple passages while retaining their multipotent differentiation capacity . They are therefore an attractive cellular source for tissue-engineering applications [2–4]. Under permissive stimulation, MSCs undergo osteogenic differentiation through a well-defined pathway [5, 6], acquiring osteoblastic markers and secreting extracellular matrix (ECM) and calcium crystals [1, 7]. MSCs have been shown to be nonimmunogenic in both in vitro [8–10] and in vivo [11, 12] transplantation paradigms, suggesting their utility for both autologous and allogeneic tissue engineering applications [13–15]. Investigations into their use in various animal models have demonstrated efficacy in healing critical size calvarial defects in mice  and femoral defects in rat  and sheep  models. However, clinical translation is hampered by the low frequency at which MSCs exist in BM, especially in older age groups, in which fractures and nonunion predominate. In addition, adult BM-derived MSCs have high cellular senescence and limited proliferation capacity [7, 19] and osteogenic potential [20–22].
More recently, MSCs with osteogenic potential have been isolated from a diverse range of tissue types and ontogenies, including adipose tissue  and perinatal tissues such as umbilical cord , umbilical cord blood [25, 26], amniotic fluid [27, 28], and fetal blood, bone marrow, and liver [29–32]. Although investigations into their basic biology, immunogenicity, and osteogenic potential have been reported, MSCs have not been systematically compared for bone tissue-engineering applications. Hence, it remains unclear how these novel fetal perinatal and adult MSC sources compare with their standard adult BM MSC counterparts for osteogenic differentiation and potential for tissue engineering.
Three-dimensional (3D) scaffolds provide the necessary support for cells to attach, grow, and differentiate, and define the overall shape of the tissue-engineered transplant . A range of biomaterials has been investigated for use in bone tissue-engineering scaffolds, which can be considered chiefly in two categories: bioceramic material and biodegradable polymers . Although bioceramic material, such as synthetic hydroxylapatite or β-tricalcium phosphate (β-TCP), has been shown to be osteoconductive, the use of ceramics is limited by their poor mechanical properties and difficulty in morphological processing . Recently, we explored the use of a polymer ceramic composite material for bone tissue-engineering scaffolds, which is made of poly ε-caprolactone (PCL) and bioactive ceramic β-TCP . This can be fabricated into honeycomb structures that allow for rapid vascularization as well as maintenance of the structural integrity of tissue-engineered bone grafts in load-bearing applications .
We compared four types of MSC from different ontological and anatomical origins in a direct head-to-head manner. In vitro comparative studies were done in both a monolayer culture system and on 3D PCL-TCP bioactive scaffold cultures to investigate their proliferation capacity, osteogenic differentiation and mineralization, and in vivo ectopic bone formation. We report that the ontological and anatomical origin of MSCs has a profound influence on their proliferative and osteogenic capacity, with hfMSCs being the most promising candidate for bone tissue engineering.
MATERIALS AND METHODS
Samples, Animals, and Ethics
All human tissue collection for research purposes was approved by the Domain Specific Review Board of National University Hospital, in compliance with international guidelines regarding the use of fetal tissue for research . In all cases, patients gave separate written consent for the use of the collected tissue. Fetal gestational age was determined by ultrasonic crown–rump or femur length measurements. Fetal femurs were collected for isolation of hfMSCs after clinically indicated termination of pregnancy. Samples correspond to 10+6, 11+1, 14+2, 17+0, and 18+3 weeks(+days) gestation (n = 5). Umbilical cords were collected following term deliveries (n = 3). Adipocyte-derived MSCs were derived from adipose tissue harvested during cosmetic surgery (donor ages, 20, 48, and 56 years; n = 3). Human adult MSC samples used in this study were provided by the Tulane University Health Sciences Center (donor ages, 19, 22, and 35 years; n = 3).
Male nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice were acquired through Charles Rivers, Australia, and all procedures were approved by the Institutional Animal Care and Use Committee at National University of Singapore. All materials used were purchased from Sigma-Aldrich (Singapore, http://www.sigmaaldrich.com) unless otherwise stated.
Isolation and Characterization of MSCs
Routine culture for all MSCs was conducted in Dulbecco's modified Eagle's medium (DMEM)-Glutamax (Gibco, Grand Island, NY, http://www.invitrogen.com) supplemented with 10% fetal bovine serum, 50 U/ml penicillin, and streptomycin (Gibco), hereafter referred to as D10 medium.
hfMSCs were isolated as previously described [31, 32]. Briefly, single-cell suspensions were prepared by flushing the BM cells out of femurs using a 22-gauge needle, passing through a 70-μm cell strainer (BD Biosciences, San Diego, http://www.bdbiosciences.com), and plating on Petri dishes (Nunc, Rochester, NY, http://www.nuncbrand.com) in D10 medium at 106 cells/ml. Adherent spindle-shaped cells were recovered from the primary culture after 4–7 days. Nonadherent cells were removed with initial medium changes every 2–3 days. At subconfluence, they were trypsinized and replated at low density (104 cells/cm2).
hUCMSCs were isolated in a similar manner described by others . Briefly, umbilical cords were washed with phosphate-buffered saline (Gibco), umbilical cord arteries were immersed in 1% collagenase for 20 minutes at 37°C, and cells were pelleted by centrifuging before plating in D10 medium. Emerging spindle-shaped adherent cells were cultured as above.
hATMSCs were isolated as previously described by others . Briefly, adipose tissue was washed before digestion with collagenase type I (1:1,000 w/v) for 60 minutes at 37°C with intermittent shaking. After removal of floating adipocytes by centrifugation (300 ×g for 5 minutes), the stromal-vascular fraction was plated at 3,500 cells/cm2 in D10 medium, and hATMSCs were recovered as above.
human BM-derived MSCs obtained from Tulane University were thawed and cultured as above.
All experiments were performed on passage 4 cells.
MSC samples were screened by immunocytochemistry (ICC) (all from DAKO, Glostrup, Denmark, http://www.dako.com, unless otherwise stated) and flow cytometry was performed as previously described . ICC was used to screen for CD14, CD34, CD45, CD31, von Willebrand factor (vWF) (Abcam, Cambridge, MA, http://www.abcam.com), CD105 (SH2), CD73 (SH3, SH4) (Abcam), vimentin, laminin, CD29 (Chemicon, Temecula, CA, http://www.chemicon.com), CD44 (BD Biosciences), CD106, CD90 (Chemicon), Oct-4 (Abcam), and Nanog (Abcam), whereas flow cytometry was used to search for HLA-I, HLA-II, and Stro-1 (Chemicon).
For osteogenic induction, MSCs were plated at 2 × 104 cells/cm2 and cultured in osteogenic differentiation medium (D10 medium supplemented with 10 mM β-glycerophosphate, 10−8 M dexamethasone, and 0.2 mM ascorbic acid) for up to 20 days, with medium changed three times per week. Extracellular accumulation of calcium was assayed by von Kossa staining. For adipogenic induction, MSCs were plated at 2 × 104 cells/cm2 and cultured in adipogenic differentiation medium (D10 medium supplemented with 5 μg/ml insulin, 10−6 M dexamethasone, and 0.6 × 10−4 M indomethacin) for up to 5 weeks with medium changed three times per week. The existence of lipid vacuoles was confirmed by oil red O staining. For chondrogenic induction, MSCs were pelleted and cultured in chondrogenic differentiation medium (DMEM supplemented with 0.1 μM dexamethasone, 0.17 mM ascorbic acid, 1.0 mM sodium pyruvate, 0.35 mM L-proline, 1% insulin-transferrin sodium-selenite (ThermoFisher Scientific, Singapore, http://www.fishersci.com), 1.25 mg/ml bovine serum albumin, 5.33 μg/ml linoleic acid, and 0.01 μg/ml transforming growth factor-β) for 28 days with medium changed three times per week. The micromass pellets were formalin fixed, paraffin embedded, and sectioned in 10-μm slices. Thereafter, they were dewaxed and rehydrated before safranin O staining .
Growth Kinetics and Colony-Forming Unit–Fibroblast Assay
The growth kinetics of MSCs were assessed by plating cells at 104 cells/cm2 in D10 medium in triplicate. Cells were trypsinized every 3 days, their numbers were enumerated, and population doubling times were calculated. The colony-forming unit–fibroblast (CFU-F) capacity of MSCs was assessed by low-density plating of MSCs at 4 cells/cm2 in 100-mm dishes (200 cells per dish) in D10 medium for 14 days, and staining with 3% crystal violet in 100% methanol for 5 minutes at room temperature. Colonies ≥2 mm in diameter were counted.
MSCs were plated for osteogenic differentiation as above for up to 28 days, with medium changed three times per week. Samples were harvested in triplicate for the following assays. The calcium content assay was done by dissolving crystals with 0.4 ml 0.5 N acetic acid overnight, and quantifying them with a calcium assay kit (BioAssay Systems, Hayward, CA, http://www.bioassaysys.com) according to the manufacturer's instructions. The amount of calcium released from acellular scaffolds grown in osteogenic medium was 23.5 ± 3.7, 27.8 ± 3.1, 35.9 ± 7.8, and 36.7 ± 5.4 μg/scaffold at days 7, 14, 21, and 28, respectively. This was then subtracted from the values obtained in the test groups. The alkaline phosphatase (ALP) activity in cell lysates was measured using SensoLyte pNPP Alkaline Phosphatase Assay Kit (AnaSpec, San Jose, CA, http://www.anaspec.com) following the manufacturer's instructions and normalized to total protein content through the Bradford assay (Bio-Rad, Hercules, CA, http://www.bio-rad.com).
Scaffold Manufacturing and Loading of Cells in Scaffolds
PCL-TCP 3D bioactive scaffold specimens were fabricated as reported by others. We chose a lay-down pattern of 0°/60°/120° to give a honeycomb-like pattern of triangular pores with a porosity of 70% and average pore size of 0.523 mm. Scaffolds 4 mm in height and 5 mm diameter were generated with an internal empty space of 55 mm3 per scaffold for cellular loading. The surface was treated with 5 M NaOH for 3 hours to enhance hydrophilicity , and all scaffolds were sterilized before cellular loading.
MSCs were suspended in fibrin glue (Tisseel kit, Immuno AG, Vienna, Austria, http://www.baxter.com) for seeding into the porous scaffolds (1.6 × 105 cells per scaffold, 3,000 cells/mm3). Cultures were maintained in 24-well plates with D10 medium and incubated overnight before transferring to osteogenic medium and culturing in osteogenic medium for 4 weeks with medium changes thrice weekly.
Cellular Adhesion, Viability, and Proliferation in 3D Scaffold Culture
Cellular morphology, adhesion, and ECM production were examined daily by phase contrast light microscopy over 28 days. Analysis of cell viability was performed with fluorescein diacetate/propidium iodide (FDA/PI) staining in triplicate, where FDA stains viable cells green and PI stains dead cells red, as previously described . Samples were imaged using a confocal laser scanning microscope (Olympus FV300 Fluoview, Olympus, Tokyo, http://www.olympus-global.com).
Total cell numbers in 3D cellular-scaffold constructs on days 1, 4, 7, 14, and 28 (n = 3, separate scaffolds from those used for FDA/PI) were estimated by quantifying the DNA content of each scaffold using the PicoGreen DNA Quantification Kit (Molecular Probes Inc., Eugene, OR, http://probes.invitrogen.com) as per the manufacturer's instructions.
Osteogenic Differentiation and Mineralization Assays in 3D Scaffold Culture
Osteogenic Gene Expression.
Cellular-scaffold constructs were collected on days 1, 7, 14, 21, 28 for RNA extraction and conversion to cDNA as previously described . We undertook real-time Taqman reverse transcription-polymerase chain reaction (RT-PCR) of the following osteogenic genes: collagen 1A1 (COL1A1), osteonectin, RunX2, and ALP (Table 1). PCRs were performed in triplicate, in 25 μl: 5 μl cDNA, 12.5 μl TaqMan Universal PCR Master Mix (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com), and 7.5 μl primer working solution. Thermal cycle conditions were 50°C for 2 minutes, 95°C for 10 minutes, then 50 cycles at 95°C for 15 seconds and 60°C for 1 minute. Amplifications were monitored with the ABI Prism 7000 Sequence Detection System (Applied Biosystems). Results were normalized against the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH), and relative gene expression was analyzed with the 2−ddCt method.
Micro-CT was performed with the 1076 SkyScan machine (SkyScan, Kontich, Belgium, http://www.skyscan.be). Cellular-scaffold constructs (n = 3 from each group) were fixed in 2.5% gluteraldehyde and placed in the sample holder for scanning through 180° with a rotation step of 1° at a spatial resolution of 35 μm. An averaging of five and a 1-mm aluminum filter were used during scanning. Scan files were reconstructed at a step size of one using a modified Feldkamp algorithm as provided by SkyScan. Reconstructed data were loaded into 3D modeling software, VGstudio (Volume Graphics GmbH, Heidelberg, Germany, http://www.volumegraphics.com), to stack the two-dimensional (2D) image into a 3D model for quantitative histomorphometric analysis. The threshold used was 200. Control acellular scaffolds that had been cultured in osteogenic medium over the same period of time were imaged, and the amount of mineralization (0.065 ± 0.023 mm3) was subtracted from the measurements made in the test groups. During scanning the new bone formation in vivo, acellular scaffold implants were used as negative controls for the subtraction as well.
Scanning Electron Microscope (SEM) and Energy Dispersive X-ray (EDX) Spectrometer Analysis.
Fixed scaffolds (n = 3 from each group) were then dehydrated in a graded ethanol series, air-dried, and gold sputtered with SCD 005 gold sputter machine (Bal-Tec AG, Balzers, Liechtenstein, http://www.bal-tec.com) for 70 seconds at 30 mA under high vacuum. The samples were viewed with a JSM-6700 SEM with an EDX accessory (JEOL Ltd., Peabody, MA, http://www.jeol.com) operating at 10 kV under high vacuum mode. The elemental composition of the nodules inside the samples was analyzed by EDX.
In Vivo Transplantation and Ectopic Bone Assays
Cellular-Scaffold Construct Preparation.
MSCs were seeded onto PCL-TCP scaffolds and predifferentiated in osteogenic differentiation medium for 2 weeks before implantation. Surgical procedure: After inducing general anesthesia, a midline longitudinal skin incision was made on the dorsal surface of each mouse, and subcutaneous pockets were created, into which the MSC cellular-scaffold constructs were inserted. The skin was closed with interrupted 6-O vicryl sutures. After 2 months, animals were euthanized and the implants were retrieved for histological and micro-CT analysis in triplicate.
Cellular-scaffold constructs (n = 3) from each group were embedded in OCT medium (Tissue-Tek; Sakura Finetek, Tokyo, Japan, http://www.sakura-finetek.com) and sectioned at a 30-μm thickness with a cryostat (CM 3050S; Leica Microsystems GmbH, Wetzlar, Germany, http://www.leica-microsystems.com). Sections were stained with von Kossa and counterstained with hematoxylin and eosin (H&E) to visualize tissue morphology and evidence of new bone formation.
Lamin A/C immunostaining was used to investigate chimerism of human cells in murine tissue as previously described [31, 32]. Briefly, sections from each sample were blocked with 5% normal goat serum for 2 hours and left to react with monoclonal mouse anti-human Lamin A/C antibody (1:100; Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com) overnight; sections were then incubated with goat anti-mouse secondary antibodies (1:100, Alexa Fluor® 488; Invitrogen, Renfrew, U.K.; http://www.invitrogen.com) for 1 hour and counterstained with PI. Images were visualized through confocal microscopy as above. The number of human and murine cells within scaffolds (n = 3) was enumerated manually for six low-powered fields (LPFs) to calculate the rate of chimerism of human cells. In total, 1,189 cells (range, 133–254 per LPF; mean, 115) were counted for each specimen.
Parametric data are shown as mean ± standard deviation, and were analyzed using one-way and two-way analysis of variance. p < .05 was considered significant.
Characterization of MSCs from Various Ontological and Anatomical Sites
hfMSCs, haMSCs, hUCMSCs, and hATMSCs shared a similar spindle-shaped morphology when cultured in monolayers (Fig. 1 A). ICC staining and flow cytometry revealed a consistent immunophenotype that was negative for hemopoietic (CD14, CD34, CD45) and endothelial (CD31, vWF) markers, and positive for mesenchymal markers (CD105 [SH2], CD73 [SH3, SH4]), intracellular markers (vimentin and laminin), and cell adhesion molecules (CD29, CD44, CD106, CD90). They expressed HLA-I but not HLA-II, with hfMSCs having a lower expression level of HLA-I (55.0%) than the other MSC types (95.5%–99.1%; Fig. 1B). hfMSCs had a higher expression level of Stro-1 (51.2%) than the other MSCs (10.4%–27.2%). In addition, gating revealed a subpopulation of hfMSCs (37.1% and 24.3%, respectively) and hUCMSCs (32.5% and 20.4%, respectively) that expressed the embryonic stem cell markers Oct-4 and Nanog, both of which were absent in haMSCs and hATMSCs (Fig. 1B).
All MSC sources differentiated readily into osteogenic, adipogenic, and chondrogenic lineages under permissive conditions. Exposure to osteogenic inductive medium resulted in secretion of extracellular calcium crystals, identified on von Kossa staining, indicating osteogenic differentiation (Fig. 1C). When cultured in adipogenic inductive medium, intracytoplasmic lipid vacuoles were observed from day 14, confirmed by oil red O staining (Fig. 1C). After 30 days of cell pellet culture in chondrogenic inductive medium, chondrogenic differentiation was observed, with hfMSCs developing the largest cell pellet, followed by haMSCs, whereas both hUCMSCs and hATMSCs displayed much less safranin O staining (Fig. 1C).
hfMSCs Have Higher Proliferative and Osteogenic Potential than hUCMSCs, haMSCs, and hATMSCs in Monolayer Cultures
In 2D monolayer cultures, hfMSCs proliferated the fastest (population doubling time, 32.3 ± 2.5 hours; n = 3), followed by hUCMSCs (54.7 ± 4.3 hours; n = 3) and hATMSCs (70.4 ± 3.6 hours; n = 3), with the slowest being haMSCs (116.6 ± 22.4 hours; n = 3), resulting in a significant difference in the number of population doublings achieved over the 18-day culture period (p < .01; Fig. 2 A). In addition, hfMSCs had a higher CFU-F forming ability, with 75.1% ± 5.0% of cells forming colonies, compared with haMSCs (39.6% ± 4.6%), hUCMSCs (47.5% ± 7.5%), and hATMSCs (37.5% ± 5.6%) (p < .01; Fig. 2B).
When exposed to osteogenic medium over 28 days, hfMSCs underwent more robust osteogenic differentiation, generating more extracellular mineralization than the other MSC types, as shown by more intense von Kossa staining (Fig. 2C). This was confirmed with direct quantification of calcium deposition, showing a 1.6- to 2.7-fold increase (Fig. 2D; p < .01) in ALP activity, indicating a 5.1- to 12.4-fold increase over the other MSC types (Fig. 2E; p < .01). Overall, the order of preferential osteogenic differentiation was hfMSCs, followed by hUCMSCs, haMSCs, and finally hATMSCs (Fig. 2C–2E).
hfMSC PCL-TCP Cellular-Scaffold Constructs Demonstrated the Highest Proliferation Capacity
Next, we loaded the various MSCs onto high-porosity PCL-TCP scaffolds (Fig. 3 A) and cultured them over 28 days to test their proliferative and osteogenic capacities in a 3D culture paradigm. FDA/PI staining of the cellular-scaffold constructs demonstrated high cellular viability over 28 days in culture in all four MSC constructs (Fig. 3B). After loading with fibrin glue, all four MSC types remained spherical during the initial period, due to their confinement within the fibrin gel. Over the next 7–14 days, they assumed a spindle-shaped morphology reminiscent of that seen in monolayer cultures, with hfMSCs achieving this at an earlier time point, at day 7, compared with day 14 for the other MSC types (Fig. 3B).
hfMSCs reached confluence within the scaffold, taking up all available spaces by day 7, whereas the other MSC types achieved confluence only at the end of the 28-day experimental period (Fig. 3B). This was confirmed by an analysis of cell numbers with a double-stranded DNA (dsDNA) quantification method, showing a sharp increase during the first week of culture in hfMSCs, undergoing a fourfold further increase between day 4 and day 7 (Fig. 3C). Thereafter, levels plateaued in hfMSC constructs, in line with the FDA/PI results. In contrast, in other MSC scaffolds, there was a small drop in dsDNA quantity from day 1 to day 4, followed by a slow increase in the dsDNA amount, which finally plateaued on day 28. hfMSCs consistently showed a higher dsDNA percentage than hUCMSCs and haMSCs from day 4 and hATMSCs from day 7 to day 28 (p < .001); hATMSC dsDNA was higher than haMSC and hUCMSC dsDNA from day 7 to day 28 (p < .001), and hUCMSC dsDNA was higher than haMSC dsDNA from day 21 onwards (p < .05; Fig. 3C).
hfMSC and haMSC Scaffolds Demonstrate Higher Osteogenic Differentiation Capacity than hUCMSC and hATMSC Scaffolds
When cultured in 3D, all four osteogenic genes were upregulated earlier (ALP) or more robustly (RunX2, COL1A1, and osteonectin) in hfMSC scaffolds than in the other MSC scaffolds. haMSCs demonstrated earlier expression of osteonectin and higher expression of ALP than hUCMSCs and hATMSCs (Fig. 4).
Using serial light microscopy, crystalline deposits were seen appearing within the scaffolds, which were most abundant in the hfMSC scaffolds, followed by the haMSC, hUCMSC, and hATMSC scaffolds when exposed to osteogenic induction medium (day 19 analysis; Fig. 5 A). In contrast, no crystalline deposits were observed in MSC scaffolds cultured in control medium. The nature of these calcium crystals was confirmed through von Kossa staining, demonstrating the darkest staining in hfMSC scaffolds followed by haMSC, hUCMSC, and hATMSC scaffolds, respectively (Fig. 5A).
Analysis through SEM of dehydrated scaffolds demonstrated a trabecular bone-like structure in all, with hfMSC scaffolds having the most extensive trabecular network of ECM, followed by haMSC scaffolds and the other two scaffolds. There was a large number of nodules found within the ECM of hfMSC scaffolds but not the other MSC scaffolds, confirmed as calcium phosphate nodules through EDX analysis (Fig. 5B).
Next, micro-CT was used to quantify mineralization of the scaffolds, allowing assessment of the entire volume of the construct. hfMSC scaffolds resulted in a 3.9- to 17.6-fold higher mineral content than the other scaffolds (p < .01), whereas haMSCs demonstrated a 4.1- and 4.5-fold higher mineral content than hUCMSCs and hATMSCs, respectively (p < .05; Fig. 5C). This finding was replicated by direct measurement of the calcium content within scaffolds through dissolution of the minerals and quantification of calcium ions. From day 14 of culture on, hfMSC scaffolds had a calcium content 1.6- to 5.0-fold higher than the other constructs (p < .01), with haMSCs demonstrating a 2.8- and 3.2-fold higher calcium content than hUCMSCs and hATMSCs, respectively (p < .01; Fig. 5D).
hfMSC and haMSC Scaffolds Demonstrate More Ectopic Bone Formation than hUCMSC and hATMSC Constructs After 2 months of Subcutaneous Implantation
Next, subcutaneous implantation in immunodeficient NOD/SCID mice was performed to compare the osteogenic potential of MSC scaffolds in vivo. Scaffolds were cultured for 2 weeks in vitro as osteogenic preinduction before implantation, and removed after 2 months for analysis. All scaffolds demonstrated neovascularization, with blood vessels infiltrating the scaffolds from the surrounding tissue macroscopically (data not shown).
In all constructs, human cells were detected in high numbers (60%–67% chimerism), as demonstrated by human-specific nuclear stain (lamins A and C), with infiltration of murine cells accounting for one third of the cellular population within the internal spaces of the scaffolds (Fig. 6 A).
Ectopic bone formation within constructs was evaluated using both traditional von Kossa histological staining and micro-CT. von Kossa staining demonstrated that hfMSC scaffolds had the largest mineralization area, followed by haMSC, hUCMSC, and hATMSC scaffolds (Fig. 6B, with H&E counterstain). Micro-CT quantification of mineralization showed that hfMSC scaffolds generated the highest bone volume (16.6 ± 3.0 mm3; p < .01), followed by haMSC scaffolds (9.1 ± 1.1 mm3; p < .05), whereas hUCMSC and hATMSC scaffolds generated the lowest new bone volume (2.9 ± 1.0 and 1.3 ± 0.1 mm3, respectively).
Large bone defects are a major clinical problem and an area of unmet need, with about one million cases requiring bone grafting in the U.S. annually [34, 42]. However, autologous grafts are not available in up to 40% of patients , and there is thus a pressing need for effective tissue-engineered solutions. Bone tissue engineering requires a porous biodegradable scaffold and a nonimmunogenic cellular source with osteogenic potential. The identification of various MSC types from different ontological and anatomical sites has raised the question as to the optimal cellular source for such allogeneic applications. Here, we compared four well-characterized MSC types from different tissues for their osteogenic potential for tissue engineering using a battery of stringent tests in the same setting. Our key finding was that hfMSCs are a superior cellular candidate not only due to their primitiveness and expression of embryonic stem cell markers, lower HLA-I expression, and higher proliferative capacities, but in particular because of their osteogenic potential both in vitro and in vivo for ectopic bone formation when compared with hUCMSCs, haMSCs, and hATMSCs.
The MSC types here were subjected to identical isolation and culture conditions in the same laboratory, to better control for the inherent differences that characterize samples processed in different laboratories. They were fully characterized in excess of criteria laid down by the International Society for Cellular Therapy , being fibroblastic in morphology, demonstrating clonogenicity, expressing various mesenchymal markers and adhesion molecules, lacking in hemopoietic and endothelial markers, and being capable of trilineage differentiation under permissive conditions. The osteogenic potential of these MSC types was investigated both in standard monolayer cultures and after loading onto an advanced generation bioactive scaffold suitable for clinical application, with results corroborated through multiple proliferation assays (cell enumeration, CFU-F capacity, FDA/PI confocal imaging, dsDNA quantification), osteogenic induction (real-time RT-PCR of key osteogenic genes and ALP activity), and ECM deposition (von Kossa staining, quantitative calcium content, micro-CT quantification, and SEM/EDX analysis). We did not perform these experiments on clonal cultures because the heterogeneity between clones makes comparisons more variable, and their utility in the clinic is limited by the immense cost involved. Moreover, nonclonal cell cultures are more likely to be clinically relevant.
Only hfMSCs and hUCMSCs expressed the pluripotency markers Oct-4 and Nanog, which are essential factors for the maintenance of pluripotency and proliferative capacity in embryonic stem cells , reflecting their primitiveness. This may, in part, explain their higher proliferative potential and CFU capacity compared with other sources, which confers advantages for rapid expansion and consequent downstream application. In addition to first trimester-derived hfMSCs , we have now demonstrated that similar markers are expressed in second trimester BM-derived hfMSCs. Oct-4+ and Nanog+ MSCs from human umbilical cord veins have also been reported by Kermani et al. , suggesting that these markers are an ontological rather than an anatomical feature. We did not find these markers in adult MSC types, although Pochampally and coworkers showed that pluripotency markers can be induced in haMSCs after a period of selection under serum deprivation .
hfMSCs had the lowest HLA-I expression, which for allogeneic applications should render them immunologically advantageous over the other MSC types. Further advantages include their lack of intracellular HLA class II and slower upregulation in response to stimulation by γ-interferon, as previously reported by Gotherstrom et al. [48–50]. Already, in cellular transplantation, allogeneic fetal hfMSC transplantation in a clinical case of osteogenesis imperfecta (OI) resulted in chimerism of up to 7% in bone, with possible clinical benefit , whereas intrauterine xenotransplantation of hfMSCs in a murine model of OI resulted in similar chimerism rates, and amelioration of the phenotype without evidence of immune rejection .
Stro-1, which is a marker most commonly associated with the osteogenic progenitor fraction found within MSC cultures [53, 54], is expressed highly in hfMSCs. This may, in part, explain their observed superior osteogenicity in monolayer and 3D scaffold cultures and in vivo ectopic bone formation over the other MSC types.
Our observation that ontogeny is an important determinant for proliferative and osteogenic capacity mirror the findings of Kim et al.  of a diminishing capacity from fetal through neonatal, infant, and juvenile BM-derived rhesus macaque MSCs, although they did not examine adult BM. More recently, Guillot et al.  demonstrated that first trimester hfMSC sources had higher expression levels of osteogenic genes, both basally and during osteogenic induction, than haMSCs in both in vitro and in vivo paradigms. A comparison of hUCMSCs and haMSCs by Baksh et al.  concurs with our findings that hUCMSCs have a higher proliferative and osteogenic differentiation capacity than haMSCs in 2D cultures, which suggests that ontogeny may play a bigger role than the anatomical origins of MSCs. Conversely, in 3D scaffold culture both in vitro and in vivo, haMSCs produced higher proliferation rates and more robust osteogenic differentiation. Although the reason for this disparity between 2D and 3D behavior for proliferation and osteogenic differentiation in haMSCs and hUCMSCs is unknown, we speculate that haMSCs may respond differently to cues associated with the adoption of a more native morphology and multidimensional cell–cell signaling processes [58–60].
We find that the anatomical origin of cultured expanded MSCs influences their osteogenic potential, with BM-derived hfMSCs and haMSCs performing better than MSCs derived from the umbilical cord and adipose tissues. This is in agreement with the findings of Im et al.  and Liu et al.  that hATMSCs have higher adipogenic but less osteogenic and chondrogenic potential than haMSCs, which has a role in producing bone and cartilage tissues in the bone marrow. Gene expression studies support this anatomical relationship to the osteogenic potential of each MSC source, with Panepucci et al.  demonstrating that haMSCs are more committed to osteogenesis, whereas hUCMSCs are more committed to angiogenesis, in keeping with their anatomical site of origin, and Guillot et al.  showing higher osteogenic gene expression in first trimester fetal BM-derived hfMSCs than in fetal blood and fetal liver hfMSCs.
We chose a panel of genes that are known to be upregulated in early (osteonectin and RunX2), intermediate (ALP), and late (COL1A1) stages of osteogenesis, and demonstrated that hfMSC scaffolds undergo earlier and more robust upregulation, mirroring their phenotypic performance in mineralization.
Scaffolds offer a 3D framework in which a temporary matrix for cellular proliferation, differentiation, and deposition of ECM allows neovasculature to develop [34, 64]. We used a bioactive PCL-TCP scaffold of high porosity to demonstrate their biocompatibility, with high cellular viability in all four MSC types after 4 weeks of in vitro culture, and 2 months on in vivo scaffolds. In addition, it allowed the deposition of trabecular bone-like ECM and calcium phosphate nodules in vitro, and ectopic bone formation within an in vivo environment.
In conclusion, the ontological and anatomical origins of MSCs have a profound influence on their proliferation and differentiation capacities, and hence affect their performance as cellular sources for bone tissue engineering. hfMSCs are the best cellular candidate, compared with hUCMSCs, haMSCs, and hATMSCs, because they express pluripotency markers and have lower immunogenicity and greater proliferative capacity. In addition, when cultured in osteogenic conditions, hfMSCs exhibited the most robust osteogenic gene induction, extracellular mineralization, and in vivo ectopic bone formation. Our data here support the use of hfMSCs as a superior allogeneic cellular source over the other MSC types for bone tissue-engineering applications. Transplantation of these cellular scaffold constructs into preclinical critical-sized femoral defect models is currently underway.
We acknowledge the following people for their kind assistance with this project: Lay Geok Tan, Praveen Vijayakumar, Yiping Fan, and Sherry Ho from the Department of Obstetrics & Gynaecology; Eddy Lee, Bina Rai, and Fenghao Chen from the Graduate Programme in Bioengineering; Erin Teo from the department of Mechanical Engineering; and Lam Xu Fu and Evelyn Susanto from the Tissue Engineering lab, National University of Singapore. We thank Citra Mattar for reviewing this manuscript. Some of the materials employed in this work were provided by the Tulane Center for Gene Therapy through a grant from NCRR of the NIH, Grant # P40RR017447.
This work is supported by grants from the National Medical Research Council (NMRC/0974/2005), the Cross Faculty Grant of NUS, Grant # R-174-000-107-123, and National Healthcare Group SIG Grant 06013 and Grant 08031, and funding from the Clinician Scientist Unit, NLAM, NUS. J.C. received salary support from an Exxon-Mobil-NUS Fellowship.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
Swee-Hin Teoh owned stock in and served as an officer or member of the board for Osteopore International.