Author contributions: M.R.J: conception and design, collection and/or assembly of data, data analysis and interpretation, and manuscript writing; W.A: collection and/or assembly of data and data analysis and interpretation; W.K.M.: collection and/or assembly of data; W.M.: data analysis and interpretation; S.R.L.: provision of study material or patients; D.G.P.: conception and design and data analysis and interpretation; O.B.F.J.: financial support, provision of study material, conception and design, manuscript writing, data analysis and interpretation, and final approval of the manuscript.
Skeletogenesis is initiated during fetal development and persists through adult life as either a remodeling process in response to homeostatic regulation or as a regenerative process in response to physical injury. Mesenchymal stem cells (MSCs) play a crucial role providing progenitor cells from which osteoblasts, bone matrix forming cells are differentiated. The mechanical environment plays an important role in regulating stem cell differentiation into osteoblasts, however, the mechanisms by which MSCs respond to mechanical stimuli are yet to be fully elucidated. To increase understanding of MSC mechanotransuction and osteogenic differentiation, this study aimed to identify novel, mechanically augmented genes and pathways with pro-osteogenic functionality. Using collagen glycoaminoglycan scaffolds as mimics of native extracellular matrix, to create a 3D environment more representative of that found in bone, MSC-seeded constructs were mechanically stimulated in a flow-perfusion bioreactor. Global gene expression profiling techniques were used to identify potential candidates warranting further investigation. Of these, placental growth factor (PGF) was selected and expression levels were shown to strongly correlate to both the magnitude and duration of mechanical stimulation. We demonstrated that PGF gene expression was modulated through an actin polymerization-mediated mechanism. The functional role of PGF in modulating MSC osteogenic differentiation was interrogated, and we showed a concentration-dependent response whereby low concentrations exhibited the strongest pro-osteogenic effect. Furthermore, pre-osteoclast migration and differentiation, as well as endothelial cell tubule formation also maintained concentration-dependent responses to PGF, suggesting a potential role for PGF in bone resorption and angiogenesis, processes key to bone remodeling and fracture repair. Stem Cells2013;31:2420–2431
Skeletogenesis is initiated during fetal development and persists through adult life as either a remodeling process in response to homeostatic regulation or as a regenerative process in response to physical injury. Bone formation can occur through intramembranous or endochondral ossification, with numerous aspects of the molecular mechanisms controlling both processes mutually expressed during fetal and adult development [1, 2]. During adulthood, however, inflammation and mechanical loading also play significant roles in orchestrating bone formation. The importance of mechanical stimuli in the normal formation, development, and homeostasis of bone is clearly observed when physiologically relevant levels of mechanical stimuli are removed [3-5]. It is hypothesized that under normal physiological conditions this dynamic remodeling of bone in response to mechanical loading occurs in order to minimize stress or strain gradients  within bone and provide an optimally mechanical structure. The role of mechanical loading in trauma or pathological conditions, such as fracture repair, is however still being fully elucidated. Traditional clinical practice dictates that fractures be stabilized in order to mediate accelerated healing, as excessive macro-motions and hence mechanical stimulation is detrimental. Conversely, alternative strategies suggest that the application of controlled “micro-motions,” during the early stages of fracture repair, beneficially affect fracture healing [7-13]. Differences observed in fracture healing, as a function of mechanical loading, have been attributed to switches in the molecular mechanisms regulating bone regeneration from an intramembranous to a more endochondral-driven mechanism, with the latter yielding delays in healing times [14, 15], but not necessarily the final quality of bone regenerated. Therefore, in vitro bone tissue engineering strategies that use mesenchymal stem cells (MSCs) have sought to incorporate biomechanical stimuli such as fluid flow-induced shear stresses in an attempt to mimic the in vivo environment and generate artificial bone graft substitutes. These strategies have shown a positive enhancement of osteogenic differentiation for both murine MSCs [16-28] and human MSCs (hMSCs) [29-32] when compared to statically cultured cells. In our laboratory, we are interested in using such in vitro cultivation strategies as model systems to identify and study the role of mechanically regulated genes, in an effort to expand the knowledge space pertaining to stem cell mechanobiology and osteogenesis. MSCs have been shown to have an inherent capacity to sense mechanical stimuli through a number of proposed mechanotransductory mechanisms, including cytoskeletal deformation [33, 34], tension [35-39], and remodeling , or alternatively via primary cilia -mediated mechanisms. These mechanosensory mechanisms have been shown to modulate numerous mechanically augmented signaling pathways, such as the Ras/Mitogen-activated protein kinase (MAPK), phosphoinositide 3-kinase (PI3K)/Akt, Rho/Rock, WntB, Transforming growth factor beta (TGF-β), and extracellular signal-regulated kinase (ERK) pathways, which have recently been reviewed extensively [42, 43]. However, the mechanisms by which cells respond to mechanical stimuli are yet to be fully elucidated. Understanding these mechanically transduced molecular mechanisms or pathways, and furthermore identifying novel differentially regulated genes, will provide insight into the key regulatory factors that govern stem cell differentiation and bone remodeling during normal homeostatic regulation and/or fracture repair. The identification of novel therapeutic candidates may then be applied using regenerative medicine strategies to either help enhance osteogenesis in diseases associated with bone loss or aid accelerated healing during fracture repair.
To increase understanding of stem cell mechanotransuction and osteogenic differentiation, this study aimed to identify novel, mechanically augmented genes and pathways with pro-osteogenic functionality. Using collagen glycoaminoglycan (CG) scaffolds as mimics of native extracellular matrix to create a 3D environment more representative of that found in bone, rat MSC (rMSC)-seeded constructs were mechanically stimulated in a flow-perfusion bioreactor. Global gene expression profiling techniques were used to identify potential candidates warranting further investigation. Placental growth factor (PGF) was selected and an actin polymerization-mediated mechanism for mechanotransduction was demonstrated. The functional role of PGF in modulating rMSC osteogenic differentiation was subsequently interrogated, and a concentration-dependent response was observed with 25 ng/mL exhibiting the strongest pro-osteogenic effect. To determine whether this finding may be clinically translatable to humans, we undertook a similar evaluation of PGF on hMSCs in 2D and 3D and demonstrated a statistically significant increase in calcium deposition for a PGF concentration of 10 ng/mL in both cases. Angiogenesis and bone resorption are also key processes in bone remodeling and fracture repair. We sought to further interrogate the concentration-dependent role of PGF with respect to these functional attributes. The most established in vitro angiogenesis assay for testing angiogenic and antiangiogenic agents uses human umbilical vein endothelial cells (HUVECs) plated on Matrigel. Using this assay, PGF exhibited the strongest proangiogenic effect at a concentration of 50 ng/mL, causing statistically significant increases in tubule length and complexity. Investigations pertaining to human osteoclast function are challenging due the considerable difficulties associated with isolating and culturing these sporadic cells. The murine macrophage cell line, RAW 264.7, is commonly used as a representative model cell type. RAW 264.7 cells respond to receptor activator of nuclear factor kappa-B ligand (RANKL) stimulation and generate multinucleated osteoclasts with the hallmark characteristics of fully differentiated cells. Using these cells, we further demonstrated that PGF at a higher concentration of 50 ng/mL could significantly enhance preosteoclast chemotaxis and differentiation in the absence of RANKL. Taken together, this study has identified PGF as a novel, mechanically augmented growth factor, capable of modulating osteogenesis, angiogenesis, and osteoclastogenesis; three key processes in bone remodeling and fracture repair.
Materials and Methods
Scaffolds were fabricated in accordance with a previously developed protocol . Briefly, a CG suspension was created by blending microfibrillar bovine tendon collagen (Integra Life Sciences, Plainsboro, NJ) with chondroitin-6-sulfate sodium salt, isolated from shark cartilage (Sigma-Aldrich, Dublin, Ireland, http://www.sigmaaldrich.com/ireland.html) in 0.05 M acetic acid (Fisher Scientific Loughborough, Leicestershire, U.K., http://www.fisherscientific.com). The CG suspension was degassed under vacuum and freeze-dried (VirTis Co., Gardiner, NY, http://www.spscientific.com) to a final freezing temperature of −10°C, which yielded scaffold sheets with mean pore sizes of 325 µm . After lyophilization, scaffold sheets were dehydrothermally crosslinked at 105°C for 24 hours in a vacuum oven at 50 mTorr (VacuCell, MMM, Germany, http://www.mmm-medcenter.de).
Individual scaffold discs (diameter = 12.7 mm; depth = 3–4 mm) were punched out of the sheets and strengthened through chemical crosslinking by submersion in an aqueous solution of 14 mM N-(3-dimethylaminopropyl)-N0-ethylcarbodiimide hydrochloride (Sigma-Aldrich) and 5.5 mM N-hydroxysuccinimide (Sigma-Aldrich) for 2 hours. Scaffolds were then washed and stored in phosphate buffered saline (PBS; Sigma-Aldrich). The final scaffolds had a porosity [mt]98.5% and a compressive modulus of 1 kPa .
rMSCs (< passage 8) were isolated from the tibiae and femora of male Wistar rats under ethical approval from the Research Ethical Committee, Royal College of Surgeons in Ireland, Dublin. Cells were cultured in Dulbecco's modified Eagle's medium (DMEM) (Sigma-Aldrich) supplemented with 10% fetal bovine serum (FBS; Biosera, East Sussex, U.K.), 1% L-glutamine (Sigma-Aldrich), 2% penicillin/streptomycin (Sigma-Aldrich), 1% Glutamax (Biosciences, Dublin, Ireland), and 1% MEM nonessential amino acids (Biosciences), which from here on will be called rat growth media (ratGM).
hMSCs were obtained as a kind gift from the Regenerative Medicine Institute, NUI Galway, Ireland. All procedures were performed with informed donor consent and ethically approved by the Clinical Research Ethical Committee at University College Hospital, Galway. Cells were cultured in standard hMSC growth medium (humanGM; DMEM supplemented with 2% penicillin/streptomycin and 10% FBS).
HUVECs (ISIS, Wicklow, Ireland) were cultured in fully supplemented (SCME002-S; Millipore, Cork, Ireland, http://www.millipore.com) EndoGro Media (SCME-BM, Millipore). For experiments requiring no recombinant vascular endothelial growth factor (VEGF), this supplement was not added to the EndoGro GM.
RAW 264.7 cells (passage <8 from ATCC stock vial) (ATCC Cat#TIB-71, LGC Standards, Middlesex, U.K.) are a murine leukemic monocyte macrophage cell line that can be readily differentiated into osteoclastic-like cells exhibiting a multinucleated morphology and high Tartrate-resistant acid phosphatase (TRAP) activity when stimulated with RANKL in the absence of Macrophage colony-stimulating factor (M-CSF). RAW 264.7 cells were cultured in high glucose DMEM (Sigma-Aldrich) supplemented with 10% heat inactivated (60°C, 1 hour) FBS, and 1% penicillin/streptomycin, which from here on will be called osteoclast growth media (OCGM). All cells were maintained as subconfluent monolayers in T175 flasks under standard conditions (37°C, 5% CO2).
Flow Perfusion Bioreactor Culture
Six cell seeded constructs were cultured simultaneously in an in-house designed flow perfusion bioreactor . Each construct had a separate syringe/scaffold-chamber/reservoir system. The flow rate was controlled through the use of a programmable syringe pump. To seed scaffolds, rMSC cells were detached from flasks with Trypsin-EDTA (Sigma-Aldrich) and resuspended at a concentration of 5 × 106 cells per milliliter in ratGM. Scaffolds were seeded with a total of 1 × 106 cells. In six-well plates, 100 µL of cell suspension was added drop-wise onto the top surface of each scaffold and the plates were placed into the incubator for 15 minutes to allow for cell attachment. Scaffolds were then turned over, and the procedure was repeated. After the second incubation period, 5 mL of ratGM was added to each well, and the scaffolds were precultured statically for 3 days to allow cell attachment and migration into the scaffold . Cell-seeded constructs (n = 6) were then transferred into the bioreactor or used as static controls (in which case they were transferred to a new six-well plate) where they were cultured for a further 2 or 7 days (media feeds were conducted on days 2 and 5) in either ratGM or rat osteogenic media (ratOM); DMEM (Sigma-Aldrich) supplemented with 10% FBS (Biosera), 2% penicillin/streptomycin (Sigma-Aldrich), 100 nM dexamethasone (Sigma-Aldrich), 50 µg/mL ascorbic acid (Sigma-Aldrich), and 10 mM β-glycerophosphate (Sigma-Aldrich). Each bioreactor had a total of 46 mL of ratGM or ratOM, while static control groups were cultured in 5 mL of ratGM or ratOM. To avoid desensitization of the constructs to mechanical stimulation while in the perfusion bioreactor system, constructs were cultured using a cyclic steady flow regime consisting of 1 hour of steady flow at 1 mL/minute followed by 7 hours of steady flow at 0.05 mL/minute for the duration of the culture period . After the culture period, constructs were washed in PBS (Sigma-Aldrich) flash frozen in liquid nitrogen and stored at −80°C until analysis.
For bioreactor studies looking at the impact of the magnitude of mechanical stimulation, cell-seeded scaffolds were cultured using a steady flow regime consisting of 1 hour of steady flow at 0.05, 1, or 5 mL/minute followed by 7 hours of steady flow at 0.05 mL/minute for the duration of the culture period. These flow rates were chosen as they equate to shear stresses in the range of 0–0.09 Pa in our system , which reside well within the range (1–1,000 × 10−4 Pa) shown to activate osteogenic differentiation in 3D systems .
For studies where actin polymerization was inhibited, cytochalasin-D (Sigma-Aldrich) at a concentration of 1 µM was added to the media for the bioreactor or the relevant static controls. A 1 µM concentration of cytochalasin D was selected after conducting a series of 2D studies under static conditions over a range of concentrations. A 1 µM concentration resulted in impaired actin polymerization as viewed by phalloidin staining but did not cause cell death or detachment from the culture surface (data not shown). Additionally, this concentration has previously been used in the literature .
Analysis of PGF Supplementation on rMSC (2D) and hMSC Osteogenesis (2D and 3D)
In order to assess the functional role of PGF with respect to stem cell osteogenesis, rMSC cells (passage <8) or hMSC cells (passage <5) were detached from flasks with trypsin-EDTA (Sigma-Aldrich).
For 2D studies, cells were replated in rat or human GM at concentrations of 5 × 104 and 3 × 104 cells per well in six-well plates respectively. After 24 hours, GM was removed and treatment media (GM; OM; OM+10 ng/mL PGF; OM+25 ng/mL PGF; OM+50 ng/mL PGF) were added to the appropriate wells, and cells were cultured for 21 days for rMSC and 17 days for hMSC (cells began to peel off the plastic after 17 days of culture for hMSC and that is why they were harvested earlier).
For 3D studies, 5 × 105 hMSCs were seeded onto collagen-Glycosaminoglycan (GAG) scaffolds. In six-well plates, 100 µL of cell suspension was added drop-wise onto the top surface of each scaffold, and the plates were placed into the incubator for 15 minutes to allow for cell attachment. Scaffolds were then turned over, and the procedure was repeated. After the second incubation period, 5 mL of humanOM was added to each well, and the scaffolds were cultured statically for 21 days.
Recombinant murine PGF (rmPGF) and human PGF (rhPGF) were used, respectively, in the rat and human studies (R&D Systems, Abingdon, U.K., http://www.rndsystems.com). Media feeds were conducted every 3–4 days. The PGF concentrations used were chosen in order to be comparable with previously published literature that examined the effect of PGF on primary human osteoblasts .
For 2D studies, at the end of the study the treatment media was removed, and 500 µL of 0.5 M HCL was added to each well. Cells were manually detached using a cell scraper, and the resulting cell suspension was placed into a 1.5 mL cryovial. Wells were washed with an additional 500 µL of 0.5 M HCL, and this was added to the cryovial. For 3D studies, constructs were placed in a cryovial containing 1 mL of 0.5 M HCL. Samples were left at 4°C for 48 hours on a shaking platform. Calcium deposition was quantified using the Calcium Liquicolor kit (Stanbio Laboratories, Boerne, TX) according to the manufacturer's protocol, and calcium concentration was deduced using a standard curve.
For 3D hMSC studies, calcium quantification data were substantiated with Alizarin Red staining. Scaffolds were fixed in 4% paraformaldehyde for 1 hour, before dehydration and paraffin embedding using an automated tissue processor (Leica, ASP300). Scaffolds were wax embedded (Leica, EG1140H) prior to sectioning at 10 µm on a slide, using a microtome (Leica, RM2255). Sections were obtained from the horizontal plane, 30%–50% from the surface of the scaffold. Calcium deposits were stained with 2% Alizarin Red solution by deparaffinizing the scaffold before staining for 5 minutes. Scaffold sections were then rinsed several times with dH2O, dehydrated with acetone and xylene. Coverslips were mounted using DPX mounting medium, and images were obtained using a microscope (Eclipse 90i, Nikon) and camera (DS-Ri1, Nikon).
Analysis of the Chemoattractive and Osteoclastogenic Nature of PGF on RAW 264.7 Cells
In order to evaluate the purported role of PGF in osteoclast migration and differentiation with respect to the PGF concentrations used in this study, a migration and differentiation study were conducted. For migration studies, RAW 264.7 pre-osteoclastic cells (2 × 104 cells) were seeded in 8 µm pore hanging cell culture inserts (Millipore) and serum starved for 2 hours at 37°C, 5% CO2 in Optimem serum-free media (Invitrogen, Ireland, http://www.invitrogen.com). Cells were then incubated for 18 hours at 37°C, 5% CO2 to assess migration through the porous membrane toward OCGM containing 10, 25, or 50 ng/mL rmPGF. Following 18-hour incubation, membranes from the migration chamber were removed and fixed in 4% formaldehyde. The membrane was then stained with hematoxylin for 10 minutes before mounting on a glass slide. Pre-osteoclasts were counted from five random fields of view, and the average cell count was established as indicative of the total cell migration. All images were taken using an inverted bright-field microscope (Leica, DMIL) at ×200 magnification.
In order to assess the functional role of PGF with respect to osteoclast differentiation, RAW 264.7 pre-osteoclastic cells were plated in OCGM at a concentration of 2 × 104 cells per well in 12-well plates. After 24 hours, OCGM was removed and treatment media were added; OCGM, OCGM+ 50 ng/mL RANKL (positive control), OCGM+10 ng/mL rmPGF, OCGM+25 ng/mL rmPGF, or OCGM+50 ng/mL rmPGF to the appropriate wells, and cells were cultured for 12 days. Media feeds were conducted every 2–3 days.
TRAP activity was quantified using absorbance of a colorimetric assay that uses p-nitrophenol phosphate. Treatment media was removed from RAW 264.7 cells, and cells were washed in PBS. A total of 200 µL of 1 mM sodium citrate solution was added to each well, and cells were incubated for 1 hour at 37°C. A total of 20 µL of this cell lysis solution was then added to 50 µL of a 100 mM p-nitrophenol phosphate, 80 mM sodium tartrate, 200 mM sodium citrate, and 200 mM sodium chloride solution in a 96-well plate. This solution was incubated for 37°C for 30 minutes, and the reaction was stopped using 1 M NaOH. Absorbance was read at 405 nm to provide a relative reading of TRAP activity in the osteoclasts.
Analysis of PGF Supplementation on HUVEC Tubule Formation
PGF has traditionally been associated with a functional role in angiogenesis under pathological conditions, such as vessel formation in hind limb ischemia or tumor induced angiogenesis. This study looked to assess the potential effects of PGF at the concentrations examined in the aforementioned osteogenesis studies to determine if a concentration-dependent difference in response of different cell types to PGF exists. HUVECs were detached from flasks with trypsin-EDTA (Sigma-Aldrich) and resuspended in the appropriate treatment media (EndoGro + 5 ng/mL VEGF [positive control]; EndoGro media − VEGF [negative control]; EndoGro − VEGF + 10 ng/mL rhPGF; EndoGro − VEGF + 25 ng/mL rhPGF; EndoGro − VEGF + 50 ng/mL rhPGF). Cells were plated (1.5 × 104 cells per well) onto BD Matrigel-Basement Membrane matrix, growth factor reduced (Unitech, Dublin, Ireland), which had previously been aliquoted (300 µL per well) in a 24-well plate and allowed to solidify in an incubator at 37°C for 45 minutes prior to experimentation. Phase contrast light microscopy images of the cells at ×10 magnification were captured at 4, 8, and 24 hours with an in situ camera (LEICA DFC420C) on a Leica DMIL microscope (Leica Microsystems, Switzerland). ImageJ software  (National Institutes of Health) was used to determine the tubule length in each image (five images per well), and the average cumulative length per well was calculated. The number of junctions formed between tubules in each image (representing the complexity of the networks formed) were counted manually.
Constructs from bioreactor experiments were resuspended and homogenized in RLT lysis buffer (Qiagen, Crawley West Sussex, U.K., http://www1.qiagen.com) supplemented with β-mercaptoethanol (1% v/v) (Sigma-Aldrich) using a rotor-stator homogenizer (Omni International, Kennesaw, GA).
For 2D studies, cell lysates were generated by washing cells in PBS, detaching with trypsin-EDTA and resuspending in RLT lysis buffer (Qiagen) supplemented with β-mercaptoethanol (1% v/v). Cell lysates from both the above procedures were then placed in QI Shredder columns (Qiagen) and centrifuged at 12,000 rpm for 30-second in an Eppendorf microcentrifuge 5415R (Eppendorf, Stevenage, U.K.), and the resulting supernatant was stored at −80°C until further analysis.
DNA quantification was performed on cell lysates using a Quant-i PicoGreen dsDNA kit (Invitrogen, Biosciences) in accordance with manufacturer's instructions. Fluorescence of the samples was measured (excitation 480 nm, emission 538 nm) using a fluorescent plate reader (Varioskan Flash 100–240 V, Thermo Scientific, DE), and DNA concentration was deduced using a standard curve.
Total RNA was extracted using a RNeasy mini kit (Qiagen) according to manufacturer's instructions. The quality and concentration of the RNA was quantified by measuring absorbance at 260 nm (Nanodrop 2000, Thermo Scientific). Reverse transcription was performed on 200 ng of total RNA using the QuantiTect RT Kit (Qiagen) according to the manufacturer's instructions. RT-PCR was subsequently performed using an Eppendorf Mastercycler Realplex 4 System (Eppendorf, Hamburg, Germany) and the QuantiTect SYBR Green PCR Kit (Qiagen) according to the manufacturer's instructions with Quanti-Tect Primers (Qiagen). Results were quantified for PGF (NM_053595) via the relative quantification (ΔΔCt) method  using 18-S as the endogenous reference. All PCR reactions were conducted in duplicate for each sample.
Total RNA extraction was conducted using the RNeasy Mini Kit (Qiagen), ensuring the complete removal of DNA via an on-column digestion of DNA using the RNase-free DNase set (Qiagen). The quality and concentration of the RNA was quantified by measuring absorbance at 260 nm using a Bioanalyzer (Agilent Technologies Inc., Santa Clara, CA, http://www.agilent.com), and RNA integrity was confirmed with the Experion RNA StdSens Analysis Kit from BioRad (700–7103). The WT-Ovation Pico RNA Amplification System (Cat. # 3300-12, NuGEN, San Carlos, CA) was used to generate and amplify cDNA from 50 ng total RNA. This single-stranded cDNA was labeled overnight using the NimbleGen One-Color DNA Labeling Kit (Cat.# 06370411011) and 4 µg of Cy3-labeled double-stranded cDNA was subsequently hybridized to a 4 × 72K rat gene expression array (Cat.# A6185-00-01, Roche NimbleGen, Madison, WI) for 16 hours. Following posthybridization washing, microarrays were scanned using an Axon 4000b microarray scanner with GenePix 6.0 (Molecular Devices, Union City, CA, http://www.moleculardevices.com). The mRNA expression data were analyzed using NimbleGens NimbleScan software, Version 2.4, which applied quantile normalization to the data , and expression values were obtained using the Robust Multi-Chip Average algorithm as described by Irizarry et al. .
This global analysis quantified expression levels for 24,000 gene identifiers for each sample. Gene identifiers were removed from further analysis if [mt]70% of the samples had values below the median of the array; a value below which the gene identifiers are considered to not be expressed. This yielded 12,520 gene identifiers for further analysis. To identify those gene identifiers having the largest contribution to the variability between samples, a principal component analysis (PCA) of all samples was performed.
PCA can be used to redefine a set of possibly correlated variables into a set of linearly uncorrelated values called principle components, where each principle component accounts for the data variability that is orthogonal (uncorrelated) to the previous components. This technique helps to identify variables (genes) that contribute significantly to the variation in the data, which can then be interrogated further. In this study, PCA was conducted using MATLAB 2007b and its associated Statistics Toolbox (MathWorks, Cambridge, U.K.). Differential gene expression levels were calculated by normalizing expression levels to the average (n = 4) for cells cultured in GM under static conditions. For dimensionality reduction, a PCA based on the correlation matrix of these data was then performed. The new parameter space was visualized for the first three principal component scores where a clear discrimination between the treatment groups of (i) GM + flow perfusion, (ii) osteo media, and (iii) Osteo media + flow perfusion was observed (Fig. 1A). Greater than 80% of the variation in the dataset was accounted for by the first three principle components (Fig. 1B) confirming the dimensionality reduction was valid. The top 1,000 gene identifiers for each of the first three principle components were extracted from the data set. Of these 3,000 identifiers, 2,186 unique identifiers were identified. The PCA was then rerun using this reduced data set to confirm that the initially observed discrimination between groups was maintained (i.e., the gene identifiers accounting for variation in the data had not been excluded) (Fig. 1C).
One-way or two-way ANOVAs were conducted in conjunction with the Holm-Sidak post hoc test for pairwise comparisons using Sigmaplot Version 11.2 (Systat Software Inc., CA). A p-value less than .05 (p < .05) was considered statistically significant. Actual p-values have been stated in the text for clarity.
To assess the strength of the linear dependence between two variables the statistical function “Pearson” in Excel (Microsoft Excel 2010, Microsoft, Redmond, WA) was used to determine the Pearson product moment correlation coefficient. A strong positive correlation is considered to exist if the r2 value returned is between .5 and 1.
Identification of Mechanically Augmented Genes by Global Gene Expression Analysis
Mechanical stimulation via flow perfusion (1 mL/minute for 1 hour followed by 7 hours at 0.05 mL/minute) of rMSC-seeded collagen-GAG scaffolds in ratGM or ratOM was applied for 48 hours using a flow perfusion bioreactor. A global analysis of gene expression levels was then undertaken using a DNAmicroarray and expression levels for 24,000 gene identifiers were quantified for each sample. Gene expression levels from the microarray data were initially validated using RT-PCR against a number of key transcription factors known to control MSC differentiation as well as proteins associated with osteogenic differentiation. This confirmed that the same trends were observed between experimental groups for both mRNA measuring techniques (Supporting Information Fig. 1). Statistical tools and data filtering techniques were then applied to identify mechanically augmented genes. First, PCA was used to identify the top 2,186 unique gene identifiers contributing to variation between treatment groups (as described in the “PCA” section of the Materials and Methods). Of these 2,186 gene identifiers, 220 gene identifiers had a [mt]twofold increase in gene expression levels when cultured in ratOM in comparison to ratGM under static conditions. These genes were therefore considered to be positively regulated after 48 hours of osteogenic differentiation. Of these 220 gene identifiers, 13 genes identifiers had a further [mt]twofold increase in gene expression when cultured in ratOM in the presence of flow perfusion. The top 10 identifiers are listed in Table 1. Of the top 10 gene identifiers, 7 had functional roles localized within the cell cytoplasm or nucleus, 2 were membrane expressed proteins, and 1 was a secreted growth factor. As we were interested in studying factors that have the potential to directly enhance osteogenesis not only of the cell they are produced in but also additionally of cells in the local microenvironment, the secreted growth factor, PGF was selected for further evaluation. PGF gene expression levels from the microarray were subsequently confirmed using RT-PCR techniques, which are considered a more sensitive method of measuring mRNA levels (Fig. 2A). These results indicated that PGF gene expression, when compared to ratGM static controls, was increased approximately fourfold under static conditions in ratOM, which further increased to approximately 10-fold when mechanical stimulation by flow perfusion was applied.
Table 1. Top 10 genes augmented by mechanical stimulation in a flow perfusion bioreactor for recombinant mesenchymal stem cell-seeded collagen-Glycosaminoglycan (GAG) scaffolds after 2 days of culture in osteogenically supplemented media
Fold change (OM+Flow/OM)
Placental growth factor was selected for further evaluation.
Similar to S100 calcium-binding protein A5 (S-100D protein)
Effect of Culture Duration and Flow Magnitude on PGF Regulation
Using rMSC-seeded scaffolds and the flow regime described in the previous section, the bioreactor culture period was extended from 2 to 7 days. PGF gene expression was increased a further twofold to fivefold in all groups as a result of the extended culture duration (Fig. 2b), suggesting that the observed response was not temporal in nature but a persistent effect relative to the length of applied treatment. Furthermore, we investigated the influence of flow magnitude on PGF expression levels by adapting the peak flow rates of the flow regime. This initial flow regime, consisting of a steady flow-rate of 1 mL/minute for 1 hour followed by 7 hours at a reduced flow rate of 0.05 mL/minute, was compared to regimes having a peak flow rate of either 0.05 mL/minute or 5 mL/minute for 1 hour followed by the reduced flow rate of 0.05 mL/minute for 7 hours over a period of 48 hours. The fold change in the mean PGF gene expression levels increased from 3.85 under static conditions (0 mL/minute) to 4.62, 11.39, and 21.86 for flow magnitudes of 0.05, 1, and 5 mL/minute, respectively. The fold change in PGF gene expression exhibited a strong linear correlation (r2 =.99 based on the mean values; r2 = .91 based on individual biological replicates, n = 18) to the flow magnitude as assessed by the Pearson product-moment correlation coefficient; a statistical tool widely used to determine the strength of linear dependence between two variables. Furthermore, a statistically significant (p < .05) increase in the level of PGF gene expression was noted at 5 mL/minute (Fig. 3A) compared to cells statically cultured in osteogenic media as measured by one-way ANOVA.
It is well accepted that when cells are confronted with considerable mechanical stresses, as in this study, they modulate the organization, abundance, and structure of stress fibers, through actin polymerization mechanisms, in order to resist these forces. Such stress fibers have been shown to be capable of modulating mechanosensory mechanisms within the cell. It was therefore hypothesized that the mechanically augmented increase in PGF gene expression was mediated via cell a cytoskeletal mechanotransduction mechanism requiring the induction of actin polymerization. To confirm this, we repeated the flow studies using the initially studied flow regime consisting of a steady flow rate of 1 mL/minute for 1 hour followed by a reduced flow rate of 0.05 mL/minute for 7 hours for a total duration of 48 hours, in the presence of the mycotoxin, Cytochalasin D; a known inhibitor of actin polymerization. Cytochalasin D inhibited statistically significant (p < .05) mechanically augmented increases in PGF expression levels, returning levels to that of the static control (Fig. 3B), but had no inhibitory effect on PGF expression under static conditions. This confirmed that actin polymerization in response to mechanical forces is required for the observed increases in PGF gene expression.
Concentration-Dependent Role of PGF During rMSC and hMSC Osteogenic Differentiation
Having identified PGF as a mechanically regulated gene with expression levels proportional to the magnitude and duration of flow perfusion applied, we assessed the potential concentration-dependent autocrine role of PGF on rMSC osteogenesis in monolayer culture. rmPGF was added at concentrations of 10, 25, and 50 ng/mL to rMSC cells cultured in ratOM for 21 days. PGF supplementation resulted in a dose-dependent U-shaped response. A 25 ng/mL concentration had the greatest pro-osteogenic effect, as measured by calcium quantification, compared to ratOM alone (Fig. 4A). To assess the potential clinical translation of this finding, we repeated the study using rhPGF in conjunction with hMSCs. Studies were initially conducted in monolayer culture and then confirmed for 3D culture, where cells were seeded on the same collagen-GAG scaffolds used in the flow perfusion studies. A similar dose-dependent response was noted to that of rMSCs, however, a statistically significantly increase (2D, p < .044; 3D, p < .05) in calcium deposition was observed at the lower PGF concentration of 10 ng/mL compared to human OM alone (Fig. 4B, 4C). Higher concentrations of 25 and 50 ng/mL had no statistically significant effect (Fig. 4B). For 3D hMSC studies, calcium quantification results were substantiated by alizarin red staining (Fig. 4D). Combined these data suggest that PGF plays a role in enhancing the osteogenic differentiation of MSC via a concentration-dependent mechanism in both human and rats.
Concentration-Dependent Role of PGF on Osteoclastogenesis
Ossification by osteoblasts is rarely observed independently under physiological conditions and is usually coupled to and proceeds resorption by osteoclasts, which act together within the context of basic multicellular units (BMUs), both in remodeling and fracture repair. Thus we investigated the potential role of mechanically augmented PGF expression on osteoclast migration and TRAP activity (an indication of the cell's resorptive potential), by adding rmPGF at concentrations of 10, 25, and 50 ng/mL to a preosteoclast cell line (RAW 264.7 cells). Cell migration was significantly increased (p < .01) at the highest PGF concentration of 50 ng/mL compared to the nonsupplemented control (Fig. 5A). Furthermore, the 50 ng/mL PGF concentration resulted in a similar level of TRAP activity to the RANKL positive control and had a statistically significant increase (p < .02) compared to the nonsupplemented control (Fig. 5B). This data further implies that as well as recruiting osteoclasts to the site of production, PGF at the 50 ng/mL concentration can additionally promote TRAP activity and thus enhance the resorptive potential of the cells.
This data in combination with the osteogenic data suggests that PGF can act in a concentration-dependent fashion with higher concentrations controlling the recruitment of osteoclasts and their activation, while lower concentrations enhances the osteogenic differentiation of MSCs.
Concentration-Dependent Role of PGF on Endothelial Cell Tubule Formation
Having noted a concentration-dependent function for PGF during MSC osteogenic differentiation and osteoclast migration/differentiation, we were interested in elucidating how this related to a third key factor associated with bone remodeling or fracture repair—angiogenesis. PGF has been acknowledged as a potent inducer of angiogenesis during pathological conditions, and angiogenesis is an important part of the early response to fracture. However, a role for PGF-associated angiogenesis under normal bone modeling or remodeling conditions has not yet been described in the literature. Thus we evaluated the concentration-dependent effect of rhPGF on the ability to induce HUVEC tubule formation. rhPGF was added at concentrations of 10, 25, and 50 ng/mL to endothelial growth medium and used to culture HUVECs for 24 hours on reduced growth factor Matrigel. Images were taken at 4, 8, and 24 hours, and groups were characterized in terms of cumulative tubule length per well or complexity (the number of junctions formed between tubules). The 50 ng/mL group had a statistically significant (p < .002) increase in cumulative tubule length compared to all other groups at all time points, and statistically increased level (p < .05) of complexity compared to all other groups for the 4- and 8-hour time points (Fig. 6).
This data in combination with the MSC osteogenesis and osteoclast data suggests PGF can act in a concentration-dependent fashion with a higher concentration of 50 ng/mL enhancing the formation of vessels and aid the recruitment and activation of osteoclasts, while lower concentrations exhibit a more potent pro-osteogenic differentiation of MSCs.
Using highly porous CG scaffolds as mimics of the native extracellular matrix found in bone, stem cell-seeded constructs were used to identify novel, mechanically augmented genes and pathways with pro-osteogenic roles, thereby increasing understanding of stem cell mechanotransduction in promoting osteogenic differentiation, while also providing insight into novel therapeutic targets to aid bone regeneration. Herein we have presented evidence to support, for the first time, the findings that PGF is a mechanosensitive gene, whose expression is regulated through an actin polymerization mechanism in MSCs, resulting in PGF expression levels proportional to both the magnitude and duration of stimulation applied. Furthermore, we have demonstrated a dose-dependent mechanism of PGF by which different cell types essential in bone formation and regeneration paradigms can be controlled/activated.
Global gene expression analysis of stem cell-seeded scaffolds undergoing chemically induced osteogenic differentiation in the presence and absence of flow perfusion led to the identification of a number of mechanically augmented genes. Of the top 10 genes identified, 7 had functional roles localized within the cell cytoplasm or nucleus, 2 were membrane expressed proteins, and 1 was a secreted growth factor. Of these genes, the secreted growth factor, PGF, a VEGF homolog, was selected for further investigation based on its potential to modulate the differentiation of not only of the cell from which it was secreted but also those cells residing within the local microenvironment.
Studies that have directly assessed the role of PGF in fracture repair using knock-out animals have shown that PGF plays an important role in regulating the transition from hypertrophic cartilage to bone, either by mediating the release of nonosteoclast-specific collagenolytic matrix metalloproteinases (MMPs), such as MMP-13, to aid matrix remodeling of the fracture callus, or through the induction of osteoclast formation . Furthermore, other researchers have shown that PGF has positive chemotaxic properties for MSCs during endrochondral ossification . Blocking of the PGF receptor, Flt-1, inhibits bone formation in vivo [59, 60]. In addition to these in vivo findings, in vitro, PGF has been shown to either directly enhance markers of bone formation in osteoblasts  and murine multipotent adult progenitor cells  or induce osteoclast differentiation [61, 62] and migration [63, 64].
Induction of PGF expression in the context of bone development, regeneration, and disease has traditionally been associated with the activation of hypoxia-mediated pathways . However, herein we have shown that it may also be mediated via mechanical stimulation and interestingly that gene expression levels in MSCs correlate to the magnitude and duration of stimulation, providing a mechanism by which the concentration of PGF can be modulated to control its functional effects. This enhancement of PGF gene expression levels in response to mechanical stimulation was shown to be directly attributable to an actin polymerization-mediated mechanism. Actin polymerization can effect cell signaling through a variety of mechanisms. Actin polymerization occurring in the cell cytoplasm allows externally applied forces to be resisted through the formation of cytoskeletal filaments that are complexed to external sites for cell adhesion, which in themselves provide membrane bound complexes for integrin or RhoGTPase-mediated signaling. Furthermore, these filaments provide trafficking routes through the cytoplasm that can enhance signal transduction. Alternatively, actin monomers are known to bind to transcriptional regulators (TRs) in the cytoplasm and inhibit their nuclear localization. Thus, events that result in dissociation of this actin monomer-TR complex (e.g., mechanically induced actin polymerization), mean the TRs are separated from the actin monomers allowing them to cross into the nucleus and initiate transcription . Interestingly, NADPH oxidase 1 (NOX1), a specialized superoxide-generating enzyme complex, was also identified during our initial microarray study and exhibited a large fold change in gene expression levels in response to mechanical stimulation (Table 1). Rac1 has been shown to regulate the ratio of F-actin to monomeric actin through a NOX1-mediated mechanism . This direct regulation of the cytoskeletal organization may be one mechanism by which subsequent signaling pathways are activated. Alternatively, reactive oxygen species (ROS) generated by NOX enzymes have been shown to be capable of directly regulating stem/progenitor cell function, especially with respect to angiogenesis, where PGF has been identified as one of the potential stimulatory factors . ROS are known to be capable of activating MAPK, Akt, and ERK signaling pathways in different cell types and previous research has shown that mechanical stimulation of MSCs by shock waves could induce an ERK-dependent increase in the osteogenic transcription factor CBFA1 (RUNX2) and enhance MSC differentiation towards an osteoprogenitor phenotype , as well as increasing VEGF-A production in osteoblasts through the same pathway . Interestingly and coincidentally, amplified NOX-driven ROS production is associated with chondrocyte death and the partial regulation of metalloproteinase-mediated cartilage matrix degradation during immune complex arthritis , processes that were significantly impaired in calluses of PGF−/− mice, which in turn resulted in the persistence of the hypertrophic cartilage masses during normal fracture repair via endochondral ossification . ROS have also been shown to activate classical bone morphogenetic protein (BMP) signaling and osteogenic differentiation . Moreover, BMP2 has previously been shown to upregulate PGF  and DLX3  gene expression levels. DLX3 is a transcription factor capable of modulating osteogenic differentiation through both RUNX2-dependent and -independent mechanisms and was also one of the top 10 mechanically augmented genes identified in this study (Table 1). The precise pathway by which actin polymerization is modulating the observed effect with respect to PGF gene expression in this study, and how this may modulate the subsequent osteogenic differentiation of the MSCs, requires further investigation.
Interestingly, this study may have highlighted, and begun to provide insight, into the role of PGF as a mechanosensitive gene in MSCs that could explain results of studies such as those conducted by Fang et al.  and others [7, 13], where the degree of mechanical motion (mechanical load applied) during fracture repair, or the acuteness of distraction osteogenesis, determines the success of the clinical outcome, as well as providing insight into how aging and age-related diseases (osteoporosis), which have been associated with changes in cell mechanosensitivity [76, 77] or increased levels of mechanical loading [78, 79], influence normal bone remodeling  and the ability to heal fracture repairs [81, 82].
This study has also shown that PGF can directly act upon both rMSCs and hMSCs enhancing the level of osteogenic differentiation at lower concentrations (25 and 10 ng/mL respectively), while also providing further validation of the role in osteoclast migration and differentiation at higher PGF concentrations (50 ng/mL). By maintaining constant concentrations of PGF supplementation across both osteogenic and osteoclastogenic studies, this study has demonstrated the concentration-dependent action of PGF, which may suggest a role for PGF in regulating BMU remodeling behavior; high magnitude mechanical stimulation of MSCs causes large increases in PGF levels, thereby recruiting osteoclasts and enhancing their resorption activities, as concentration levels decrease, the MSC osteogenic differentiation is enhanced, resulting in osteoblasts that then in turn form the resulting bone matrix.
In addition to these functional roles in bone formation and resorption, PGF has been well-established as a modulator of angiogenesis: a prerequisite for both intramembranous and endochondral ossification. We confirmed herein that PGF enhances the tubule formation capacity and complexity of HUVECs in a concentration dependent manner, with a concentration of 50 ng/mL yielding statistically significant increases. Regression of the tubules, as observed between 8 and 24 hours in this study, is typical of the expected behavior of these capillary-like structures [83-85] in the absence of secondary cues or stabilizing cell types . A number of intramolecular and intermolecular mechanisms have been suggested by which PGF can regulate angiogenesis, either directly or through enhancing the functionality of VEGF [86-88]. Interestingly, inhibiting VEGF production has also been shown to impair endochondral and intramembranous healing in vivo [89, 90] and ossification in vitro . Thus changes in both VEGF and PGF expression levels have the potential to alter the dynamics of bone formation and regeneration, with PGF also offering a means by which VEGF can be modulated.
The data from this study, combined with evidence from the literature, suggest that PGF plays an important role with respect to bone remodeling and regeneration. What has been lacking is a potential explanation regarding how a single molecule can have the capacity to control and modulate such a varied set of biological processes and cell subsets. Our hypothesis is that PGF functions in a concentration-dependent fashion, with different cell types and biological processes being enhanced within different concentration ranges; osteogenesis at lower concentrations and angiogenesis/osteoclastogenesis at higher concentrations.
While studies such as these help to expand the knowledge space pertaining to stem cell mechanobiology and osteogenic differentiation, it is important to acknowledge their potential impact on regenerative medicine and bone tissue engineering strategies. Controlled release of growth factors  or genes encoding for these proteins , in conjunction with scaffolds that have been shown to have natural osteoconductive and osteoinductive properties [93, 94], for the purpose of enhancing the localized repair of bone tissue, is a rapidly expanding research area of interest, which has already shown commercial success (Medtronic INFUSE Bone Graft). However, these have typically focused on the use of BMPs, which have been shown to have undesirable long-term side effects, due to the large doses that have to be administered in order to achieve a therapeutic effect. Identification of PGF as a pro-osteogenic growth factor capable of augmenting osteogenic differentiation, at much lower concentrations than those required for BMPs, gives clinicians another option to select from as the clinical efficacy of currently identified growth factors is weighed up against other factors such as potential side effects and economical constraints.
In conclusion, this study has identified PGF as a mechanoregulated gene in MSCs whose expression level is sensitive to the magnitude and duration of stimulation. Furthermore, we have shown its secretion has the ability to act in an autocrine fashion on MSCs, enhancing osteogenic differentiation at low concentrations, while also being able to induce and modulate osteoclastogenesis and angiogenesis at higher concentrations; pre-requisites for bone remodeling and fracture repair. Taken together, these results identify for the first time a mechanosensitive gene in MSCs capable of functionally driving different arms of bone remodeling and fracture repair and supports the idea that PGF offers a potential therapeutic-target for bone related diseases where the mechanical sensitivity of MSCs are disrupted or the mechanical loading of bone has been significantly altered.
We thank Dr. Caroline Curtin for technical support and assistance in the hMSC osteogenesis assay experiments. This study has received funding from the European Research Council under the European Community's Seventh Framework Programme (FP7/2007-2013) under ERC grant agreement no. 239685. Collagen materials were provided by Integra Life Sciences, Inc. through a Material Transfer Agreement.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.