A canine in vitro model for evaluation of marrow-derived mesenchymal stromal cell-based bone scaffolds

Abstract

Tissue engineered bone grafts based on bone marrow mesenchymal stromal cells (MSCs) are being actively developed for craniomaxillofacial (CMF) applications. As for all tissue engineered implants, the bone-regenerating capacity of these MSC-based grafts must first be evaluated in animal models prior to human trials. Canine models have traditionally resulted in improved clinical translation of CMF grafts relative to other animal models. However, the utility of canine CMF models for evaluating MSC-based bone grafts rests on canine MSCs (cMSCs) responding in a similar manner to scaffold-based stimuli as human MSCs (hMSCs). Herein, cMSC and hMSC responses to polyethylene glycol (PEG)-based scaffolds were therefore compared in the presence or absence of osteoinductive polydimethylsiloxane (PDMS). Notably, the conjugation of PDMS to PEG-based constructs resulted in increases in both cMSC and hMSC osteopontin and calcium deposition. Based on these results, cMSCs were further used to assess the efficacy of tethered bone morphogenic protein 2 (BMP2) in enhancing PEG-PDMS scaffold osteoinductivity. Addition of low doses of tethered BMP2 (100 ng/mL) to PEG-PDMS systems increased cMSC expression of osterix and osteopontin compared to both PEG-PDMS and PEG-BMP2 controls. Furthermore, these increases were comparable to effects seen with up to 5-times higher BMP2 doses noted in literature.

1. INTRODUCTION

Craniomaxillofacial (CMF) defects are a significant source of trauma among combat-injured military personnel and veterans ¹. In addition, CMF defects are one of the most common birth defects ^2–4, with approximately 1 in 700 live births being associated with a cleft or palate deformity ³. Such defects are often associated with substantial pain, aesthetic concerns, and high cost of treatment, which generally includes surgical procedures, dental treatment, speech therapy, and counseling ⁵. Surgical interventions generally require the use of autografts or allografts, such as allogeneic bone or demineralized bone matrix ^4,6–9. However, each of these graft sources has inherent disadvantages, such as donor site morbidity and prolonged post-operative pain for autografts, immune rejection and disease transmission for allografts, and limited availability for both.

Tissue engineering is a promising alternative to autograft- or allograft-based treatment of CMF bone defects ^4,6,9–12. In the context of tissue engineering, scaffolds intended to support bone regeneration have been developed from numerous materials, including ceramics ^13–20, synthetic polymers ^21–23, and metals ^24–26. In many cases, these scaffolds are seeded with bone marrow mesenchymal stromal cells (MSCs) ²⁷, a multipotent cell type capable of differentiating into osteogenic, chondrogenic, and adipogenic lineages. In MSC-based bone scaffold approaches, the biomaterial is generally modified to support a strong and specific osteogenic response from associated MSCs, often through the inclusion of bioactive glass ^17,28–30, hydroxyapatite (HAp) ^13–16, or osteoinductive growth factors ^31,32.

As for all tissue engineered bone grafts, the bone-regenerating capacity of these MSC-based constructs must first be evaluated in animal models prior to human trials. Canine models have traditionally been associated with improved clinical translation of CMF grafts and devices relative to small animal models (such as rodents) ^33,34 as well as larger animal models (such as rabbit, sheep, and goat) ^34–36. This improved translation potential results in part from the greater similarity between human and canine bone remodeling, bone composition, and dentition ^35,37. However, extending canine CMF models to the evaluation of MSC-based bone scaffolds also requires a high degree of correlation in the osteogenic responses of canine MSCs (cMSCs) and human MSCs (hMSCs) to scaffold-based stimuli. Currently, limited evidence exists to support the conclusion that cMSCs respond similarly to 3D osteogenic environments as their human counterparts ³⁸. Thus, a primary goal of the current study is to compare the osteogenic responses of cMSCs and hMSCs to equivalent 3D scaffold environments.

In comparing cMSC to hMSC responses, polyethylene glycol (PEG) scaffolds as well as PEG-polydimethylsiloxane (PEG-PDMS) constructs ^39–41 were utilized. Notably, while PEG-only scaffolds are generally considered relatively inert, they have been found to support osteoblastic differentiation when coupled with silica- or siloxane- based additives such as PDMS ^39–44. Indeed, PEG-PDMS constructs have previously been shown to stimulate progenitor cells to undergo highly specific osteogenic lineage progression even in the absence of osteogenic media supplements, and the PDMS phase has been shown to intrinsically induce mineralization ^39–41. Beyond these points, the use of acrylate-derivatized PEG as the scaffold base allows for the covalent incorporation of additional stimuli, a feature which has been used extensively to increase PEG scaffold bioactivity ^45–47. Specifically, to enhance the intrinsic osteoinductive strength of PEG-PDMS scaffolds, growth factors such as recombinant bone morphogenic protein 2 (BMP2) can potentially be conjugated into the scaffold network.

BMP2 is known to play a significant role in the initial stages of bone healing as well as in the later stages of osteogenesis ^48–50, and both cMSCs and hMSCs undergo osteoblastic differentiation in response to exogenous BMP2 ^51,52. However, the large doses of BMP2 (as high as 1.5 mg/mL scaffold ⁵³) often administered to counter its rapid in vivo clearance have been associated with high treatment costs as well as significant complications, including inflammation, tumor development, and ectopic bone formation ^54–56. To reduce the levels of BMP2 needed to trigger desired cell responses, BMP2 can be covalently tethered within scaffolds ^31,32,57. In contrast to physical incorporation, this method prevents cell internalization of BMP2, allowing for repeated signaling between the cell and the ligand, and prolongs BMP2 retention within the scaffold ³¹. Indeed, 200–500 ng/mL of tethered BMP2 has previously been shown to induce MSC osteoblastic differentiation ^31,50,58,59, and 60–100 ng/mL has been reported to have an significant osteogenic effects on cMSCs ⁵¹. Thus, a second goal of the current work is to utilize validated cMSCs to assess the efficacy of tethered, low-dose BMP2 in enhancing the intrinsic osteogenic capacity of PEG-PDMS scaffolds.

Toward these ends, two sets of experiments were conducted in the current study: the first focused on comparing cMSC and hMSC osteogenic responses to equivalent 3D scaffold environments and the second focused on utilizing cMSCs to evaluate the increase in PEG-PDMS scaffold osteoinductivity following incorporation of low-dose, tethered BMP2. In each study, MSCs were cultured in various scaffold formulations in the absence of osteogenic media supplements. After 21 days of culture, levels of osteogenic transcription factors runx2 or osterix, bone ECM proteins collagen I (Col-1) and osteopontin (OPN), and mineralization were evaluated.

2. MATERIALS AND METHODS

2.1. Preparation of diacrylate-terminated PEG and methacrylate-terminated star PDMS

Diacrylate-terminated PEG (PEG-DA, M_n = 3.4 kDa) and methacrylate-terminated star PDMS (PDMS_star-MA, M_n = 2 kDa) were prepared according to previously described protocols ^39–41,60. The acrylation of PEG-DA and methacrylation of PDMS_star-MA were each confirmed to be > 95% by ¹H-NMR.

2.2. Synthesis of acrylate-derivatized RGDS and acrylate-derivatized BMP2

Acrylate-derivatization of the cell adhesion peptide RGDS (American Peptide) and recombinant human BMP2 (carrier free; R&D Systems) was achieved by reaction with acryloyl-PEG-succinimidyl valerate (ACRL-PEG-SVA, 3.4 kDa; Laysan Bio) at a 1:1 molar ratio or a 1:6 molar ratio, respectively ^61,62. Both mixtures were reacted in 50 mM sodium bicarbonate buffer, pH 8.5, for 2 h at room temperature (RT) under constant agitation. The reaction products (ACRL-PEG-RGDS) and (ACRL-PEG-BMP2) were separately purified by overnight dialysis against double deionized water (dIH₂O) using 3,500 MWCO or 5,000 MWCO Snakeskin Dialysis tubing (Thermo Scientific), respectively. ACRL-PEG-RGDS was lyophilized and stored at −20 °C until use. ACRL-PEG-BMP2 was filter-sterilized, stored at 4°C, and used within 1 h of purification.

2.3. Evaluation of scaffold material properties

Precursor solutions containing 10 wt% PEG-DA and 1 mM ACRL-PEG-RGDS were prepared in phosphate buffered saline (PBS). Photoinitiator, consisting of a 30 wt% solution of 2,2-dimethyl-2-phenyl-acetophenone in N-vinylpyrrolidone, was added to each precursor solution at a concentration of 10 µL/mL. To a subset of solutions, PDMS_star-MA was then added to 2 wt %. The mixtures were vortexed and filter sterilized using 0.22-µm polyethersulfone (PES) filters. Sterile ACRL-PEG-BMP2 was also added to a subset of precursor solutions to achieve a final BMP2 concentration of 100 ng/mL. In total, four distinct solutions were obtained (each containing 1 mM ACRL-PEG-RGDS and photoinitiator): (i) 10 wt % PEG-DA (PEG), (ii) 10 wt% PEG-DA + 2 wt% PDMS-MA (PEG-PDMS), (iii) 10 wt% PEG-DA + 100 ng/mL ACRL-PEG-BMP2 (PEG-BMP2) and (iv) 10 wt% PEG-DA + 2 wt% PDMS-MA + 100 ng/mL ACRL-PEG-BMP2 (PEG-PDMS-BMP2). These formulations were separately poured into 0.75 mm-thick, UV transparent, rectangular molds and polymerized by exposure to long-wave UV light (∼6 mW/cm², 365 nm; Spectroline) for 6 min. After polymerization, the hydrogel slabs were washed in PBS and incubated for 24 h at RT to allow the gels to reach equilibrium swelling.

2.3.1. Bulk mechanical testing

A series of 8 mm discs were prepared from the swollen hydrogel slabs using a Miltex biopsy punch. Three to four 8 mm discs were further separated into a 6 mm inner disc and an outer 8 mm ring using a 6 mm biopsy punch. The outer 8 mm rings were employed for tensile testing using a modification of the circumferential property testing technique validated by Johnson et al. ⁶³. The 6 mm inner discs were subsequently utilized for mesh size assessments. Briefly, each ring was mounted onto an Instron 3342 by threading opposing stainless steel hooks through the segment lumen. The hooks were then uniaxially stretched at a rate of 6 mm/min until construct failure. The elastic modulus, E, of each sample, was defined as the slope of the linear region of the resulting stress–strain curve.

2.3.2. Hydrogel average mesh size

In this study, hydrogel average mesh size was characterized using an equilibrium dextran diffusion assay based on an adaptation of the methodology of Watkins et al. ⁶⁴. Briefly, four 6 mm discs from each hydrogel formulation were immersed in 0.5 mL of PBS containing 50 µg/mL FITC-labeled dextran (4, 10 and 20 kDa; Sigma). Dextran was then allowed to diffuse into the hydrogels for 24 h at RT, after which each disc was gently blotted and transferred to 0.5 mL fresh PBS. After an additional 24 h at RT, the fluorescence of the PBS solution surrounding each disc was measured at ex/em 488/532 nm using a plate reader. Each fluorescence measure was converted to micrograms of dextran using dextran standard curves and then divided by gel thickness to yield a quantitative indicator of hydrogel permissivity (C). These permissivity measures were used to estimate the average mesh size (ξ_x) of each hydrogel type (x) relative to the PEG formulation as follows: ξ_x= [C_x/C_PEG].

2.4. Canine tissue collection and mesenchymal stromal cell isolation

Canine MSC isolation was conducted with the approval of the Texas A&M University Institutional Animal Care and Use Committee (Animal Use Protocol 2015-0072). Primary cMSCs were isolated from three donors to partially control for age and breed effects: 1) a 14 month old, 65 lb intact male Walker hound (∼equivalent to a 16.5 year old human), 2) a 6 year old, 148 lb neutered male mixed breed (∼equivalent to a 46–49 year old human), and 3) a 3.5 year old, 47 lb neutered male mixed breed (∼equivalent to a 30 year old human). The intact versus neutered status of the male canine donors was utilized to gain initial insight into potential gender effects, since female cMSCs were unavailable to us at the time of these studies.

Under general anesthesia, a bone marrow aspirate was performed on the craniodorsal (posterior) iliac crest of each donor using a 15 gauge Illinois biopsy needle. The craniodorsal iliac crests of two additional canine donors were also separately aspirated to allow testing of response consistency to PEG-PDMS-BMP2 scaffolds. Nucleated cells were isolated from bone marrow samples using gradient centrifugation (Ficoll-Paque Plus, GE Health Care Biosciences) using previously described techniques ⁶⁵. Following centrifugation at 1800 × g for 30 min, mononuclear cells were removed, washed twice with 15 mL of Hank’s Balanced Salt Solution (HBSS, Invitrogen), quantified and assessed for viability using a hemocytometer and trypan blue exclusion.

Nucleated marrow cells were plated at 30,000 cells/cm² in 150 mm tissue culture dishes in Complete Culture Medium (CCM) containing α-MEM, 100 units/mL penicillin and 100 µg/mL streptomycin (1% PS; Invitrogen), and 10% fetal bovine serum (FBS; Atlanta Biologicals). Cells were incubated at 37 °C and 5% humidified CO₂ for 24 h. Plates were washed with PBS to remove non-adherent cells followed by media exchange. This process was repeated daily for 2 additional days. Culture dishes were subsequently monitored for expansion of the primary cell population (passage 0) with media exchange performed every other day. Upon reaching 70% confluence (5–12 days), cells were lifted with 0.5% trypsin/EDTA solution (Invitrogen) and re-seeded at 100 cells/cm² for expansion of passage 1 (P1) cells. Cells were fed every other day until 70% confluent, after which P1 cells were cryopreserved in α-MEM with 5% DMSO (Sigma-Aldrich) and 30% FBS in preparation for subsequent experiments. For cell characterization assays, P1 cells were thawed, plated at 100 cells/cm² and expanded to 70% confluency. These cMSCs were characterized by flow cytometry for cell surface epitopes – including CD44⁺, CD105⁺, CD90⁺, CD9⁺, STRO-1⁻, CD34⁻, and CD45⁻ (AbD Serotec or R&D Systems; Supplementary Table 1) – to confirm their MSC character. Isolated cMSCs were also confirmed to be able to differentiate down adipogenic, chondrogenic, and osteogenic lineages by standard methodologies.

2.5. Human and canine mesenchymal stromal cell culture

Primary human MSCs (hMSCs) from a healthy, 22 year old female donor were obtained from Lonza (PT-2501, Lot 7F3677) at P1. These donor cells had been confirmed by Lonza to be CD44⁺, CD105⁺, CD29⁺, CD166⁺, CD14⁻, CD34⁻, and CD45⁻ and to undergo adipogenic, chondrogenic, or osteogenic differentiation under inductive culture conditions. To prepare for the encapsulation studies, the obtained P1 cMSCs and hMSCs were thawed and expanded at 37 °C and 5% CO₂.

2.6. Fabrication and culture of cell-laden constructs

For the first set of experiments, cMSCs (passage 3–4) or hMSCs (passage 3–4) were resuspended in sterile PEG or PEG-PDMS precursor solutions at 1.5×10⁶ cells/mL. The resulting suspensions were polymerized into hydrogels by exposure to long-wave UV light (∼6 mW/cm², 365 nm; Spectroline) for 6 min. Each hydrogel disc (8 mm diameter) was then rinsed with sterile PBS (pH 7.4; Invitrogen) and placed in DMEM supplemented with 10% Premium-FBS (Atlanta Biologics) and 1% PS at 37 °C and 5% CO₂. The discs were cultured for 21 days, with media changes every other day.

In the second set of experiments, cMSCs (passage 3–4) were encapsulated in PEG-PDMS, PEG-BMP2, or PEG-PDMS-BMP2 precursor solutions at 1.5×10⁶ cells/mL as described above. The hydrogel discs were cultured for 21 days in DMEM supplemented with 10% Premium-FBS and 1% PS at 37 °C and 5% CO₂, with media changes every other day.

2.7. Construct harvest

For both the first and second experiments, sets of sample discs were collected for protein analyses (n = 3–4 per formulation) and histological analyses (n = 3–4 per formulation) both at 24 h post-encapsulation (“day 0”) and following 21 days of culture. For protein-based assays, both endpoint and day 0 sample discs were flash-frozen in liquid N₂ and stored at −80 °C. Thereafter, they were homogenized in lysis buffer (Ambion, Life Technologies), and the supernatant was collected after centrifugation. A second extraction from the gel pellet was performed by conducting three freeze-thaw cycles, followed by centrifugation to isolate the supernatant, which was combined with the supernatant obtained from the initial homogenization. DNA levels in the supernatant solutions were then measured using the PicoGreen assay (Invitrogen), with calf thymus DNA (Sigma) serving as a standard. The resulting solutions were stored at −80 °C for subsequent protein analyses. For histology, collected day 0 and endpoint samples were fixed in 10% formalin for 30 min, embedded in Tissue-Tek freezing medium, and sectioned using a cryomicrotome (20 µm thickness).

2.8. Protein level analyses

For hMSC-laden samples, the levels of osteogenic markers collagen I (Col-1) and osteopontin (OPN) were analyzed using a human multiplex immunoassay kit (R&D Systems). Briefly, hydrogel homogenates were diluted 1:1 in the kit-provided buffer and dispensed into a 96-well plate, after which magnetic bead suspensions, detection antibodies, and streptavidin-phycoerythrin were added to each sample well. The plate was then loaded into MAGPIX system (xMAP technology, Luminex), and sample concentrations for each analyte were obtained on the basis of their respective median fluorescence intensities relative to standards. Resulting protein concentration measures were then subsequently normalized to sample DNA content as determined by the PicoGreen assay (Invitrogen).

For canine samples, multiplex immunoassays for the Luminex xMAP system were unavailable for the proteins of interest. As such, the levels of Col-1 and OPN associated with the various cMSC treatment groups were analyzed using Western blot immunoassays. In brief, hydrogel homogenates were concentrated using 3,000 MWCO Amicon filter units (Millipore), followed by addition of β-mercaptoethanol and heating at 95 °C for 10 min. Concentrated proteins samples with consistent DNA levels were loaded into different wells of 8 or 12% polyacrylamide gels and separated by electrophoresis. After electrophoresis, proteins were transferred to a nitrocellulose membrane (Thermo Scientific), and the membranes were blocked with a 5% bovine serum albumin (BSA; Fisher Scientific) solution in TBST/NaN₃ (25 mM Tris-HCl, pH. 7.5, 137 mM NaCl, 0.1% Tween 20, 0.05 % NaN₃) for 1 h at RT. Appropriate primary antibodies (Supplementary Table 2) were diluted in 5% BSA solution in TBST/NaN₃ and applied overnight at 4 °C with constant shaking.

Bound primary antibodies were detected by the application of appropriate horseradish peroxidase-conjugated or alkaline phosphatase-conjugated secondary antibodies (Jackson ImmunoResearch) for 1 h at RT, followed by the application of Luminol (Santa Cruz Biotechnology) or Novex chemiluminescent substrate (Life Technologies) respectively. Chemiluminescence was detected using a ChemiDoc™ XRS⁺ System equipped with Image Lab™ Software (BioRad), with exposure time controlled to avoid signal saturation. The band integrated optical density for each marker was quantified using Adobe Photoshop and normalized to the amount of loaded DNA.

For the purpose of comparison, the average DNA-normalized protein measures for each treatment group were normalized to the corresponding PEG samples (study 1) or to the corresponding PEG-PDMS controls (study 2). Following completion of the protein assessments, the hMSC and cMSC responses to PEG and PEG-PDMS scaffolds were compared. Given the strong agreement between the normalized human MAGPIX immunoassay data and the normalized canine Western blot immunoassay data, we judged it to be unnecessary to repeat the human protein measures using Western blot.

2.9. Histological analyses

For samples reserved for histology, standard von Kossa staining protocols or immunohistochemical techniques were utilized to assess MSC responses to the various construct formulations.

2.9.1. Immunohistochemical staining

For immunostaining analyses, rehydrated sections were blocked for 10 min by exposure to Terminator solution (Biocare Medical). Samples were incubated at 4 °C overnight with primary antibody diluted in PBST (PBS plus 0.1 % Tween 20) containing 3% BSA. Bound primary antibody was detected using the appropriate alkaline phosphate-conjugated secondary antibody (Jackson Immunoresearch) followed by the application of chromogen Ferangi Blue (Biocare Medical) and mounting. Stained sections were imaged using a Zeiss Axiovert microscope.

2.9.2. von Kossa staining

Extracellular calcium deposition was assessed using a von Kossa staining kit (American Mastertech). Briefly, rehydrated sections were rinsed with water, after which a 5% silver nitrate solution was applied. Sections were then exposed to full-spectrum light in a humidified chamber for 1 h. After rinsing with dIH₂O, sections were exposed to 5% sodium thiosulfate for 3 min, briefly rinsed with dIH₂O, and mounted. Stained sections were imaged using a Zeiss Axiovert microscope.

2.9.3 Semi-quantitative staining assessment

Cell counts were carried out to semi-quantitatively evaluate immunostaining results for the intracellular proteins runx2 and osterix and the cell-surface protein AFABP. In addition, since deposited ECM remained localized around the parent cells in each construct - as is characteristic for PEG-based gels ⁶⁶ - the relative levels of OPN, Col-1, and matrix mineralization (von Kossa) were also evaluated by cell counts. These counting assessments were conducted according to previously validated methods ^67–70 by two independent observers. For each cell, i, in a given section, a staining intensity, d_i, was recorded on a scale of 0–3, 0 = “no staining” and 3 = “highest intensity among all stained samples for that species”. The cumulative staining intensity, d, for a given stain in a particular section was then calculated using the following equation: d = (Σ d_i)/(total cell number). Four-to-ten sections per sample and 3–4 samples per construct type were analyzed for each stain.

2.10. Statistical analyses

All data are reported as mean ± standard error of the mean. Comparison of sample means was performed by one-way ANOVA followed by Tukey’s post-hoc test (IBM-SPSS software), with significance established at p < 0.05.

3. RESULTS

3.1. MSC characterization

The primary human MSCs (hMSCs) used in this study were obtained from Lonza and were confirmed to express CD29, CD44, CD105, and CD166 and to not express CD14, CD34, and CD45. In addition, the hMSCs were confirmed to differentiate down the adipogenic, chondrogenic, and osteogenic lineages when cultured under recommended differentiation conditions. Similarly, flow cytometry was performed to evaluate the MSC identity of the isolated canine marrow cells per previously developed marker criteria ^71,72. Specifically, canine marrow cells were evaluated for cell surface markers CD105, CD44, CD90, CD9, CD34, CD45, and STRO-1 (Figure 1A). Canine marrow cells were negative for CD34, CD45, and STRO-1, but were positive for CD105, CD90, CD44, and CD9, indicating a MSC phenotype.

Figure 1.

(A) Flow cytometry results for canine marrow stromal cells supporting the MSC character of the isolated cells. (B) Representative images for canine stromal cells cultured in osteogenic induction medium and stained with Alizarin red; canine stromal cells cultured in adipogenic induction medium and stained with Oil Red; and a canine stromal cell micromass cultures grown in chondrogenic induction medium and stained with toluidine blue. Scale bars = 100 µm. These images support the tri-lineage differentiation potential of the isolated cells.

To further assess the multipotency of the isolated cells, the canine marrow cells were cultured in osteogenic media or adipogenic media on tissue culture polystyrene for 21 days. Similarly, micromass cultures of canine marrow cells were exposed to chondrogenic media for 21 days. Figure 1B shows representative images of osteogenic, adipogenic, and chondrogenic micromass cultures stained for Alizarin Red, Oil Red O, or toluidine blue, respectively. These induction cultures demonstrate the ability of the isolated canine marrow cells to undergo osteogenic, adipogenic, and chondrogenic lineage progression. Overall, the flow cytometry results and directed differentiation cultures confirm the MSC character of the isolated canine stromal cells (hereafter referred to as cMSCs) and establish them as a viable cell source for the purposes of this study.

3.2. Canine MSC model validation

Canines are commonly used as a preclinical model for bone regeneration due to a number of similarities in pathogenesis, loading, and structure with humans ⁷³. However, extending canine data to prediction of human outcomes is complex and based on the assumption that the involved canine cells behave in a similar way to their human counterparts. For evaluation of MSC-based bone scaffolds, this remains a parameter to be considered as cMSCs present several differences relative to hMSCs, including shorter in vitro life times, lower rates of division, and distinct differentiation potentials ^74,75. We therefore compared the osteogenic responses of cMSCs and hMSCs to equivalent PEG-based scaffold environments in the presence or absence of osteoinductive PDMS.

In brief, cMSCs and hMSCs were encapsulated in PEG (10 wt%) constructs and PEG-PDMS (10 wt%, 2 wt%) constructs and cultured for 21 days. Each scaffold formulation was prepared with consistent levels of tethered RGDS to allow for cell adhesion to the scaffold network. Furthermore, characterization of the scaffolds indicated no statistical differences in the initial tensile modulus or average mesh size between the PEG and PEG-PDMS constructs (Table 1). Given the slow degradation rate of PEG and PEG-PDMS systems ⁴¹, these basal similarities allow differences in day 21 MSC responses between scaffold formulations to be attributed primarily to the presence of PDMS.

Table 1.

Comparison of the tensile modulus and relative average mesh size of the various scaffold formulations. Property results represent an average ± standard error of the mean for n = 4 samples for each construct formulation. The PEG-PDMS, PEG-BMP2 and PEG-PDMS-BMP2 formulations were included in the second phase of this study, whereas PEG versus PEG-PDMS constructs were the focus of the first phase of this work.

Experiment	Hydrogel Formulation	Tensile Modulus (kPa)	Average Mesh Size Relative to PEG Control
Study 1: cMSC-hMSC comparison	PEG	131.3 ± 1.5	1.00 ± 0.02 (∼5.0 nm 114)
	PEG-PDMS	137.8 ± 1.7	0.95 ± 0.03
Study 2: cMSC evaluation of low dose BMP2	PEG-PDMS	139.8 ± 2.2	0.97 ± 0.03
	PEG-BMP2	135.1 ± 1.7	0.99 ± 0.02
	PEG-PDMS-BMP2	138.3 ± 5.1	0.94 ± 0.01

To assess cMSC and hMSC lineage progression, the levels of the early osteogenic transcription factor runx2, ECM proteins Col-1 and OPN, and matrix mineralization were all evaluated with respect to scaffold type and relative to day 0. As shown in Figure 2, the osteogenic response of cMSCs to 21 days of culture within PEG and PEG-PDMS constructs displayed a number of similarities with hMSCs. For instance, culture of both cMSCs and hMSCs in PEG constructs resulted in substantial increases in Col-1 deposition (p < 0.004) and matrix mineralization (p < 0.006) relative to day 0. PEG-PDMS scaffolds were also associated with increased levels of calcium (p < 0.001) and OPN (p < 0.010) deposition relative to day 0 for both cell types. In comparing responses to PEG versus PEG-PDMS scaffolds, the levels of runx2, OPN, and mineralization within PEG-PDMS constructs each appeared to be greater than the corresponding PEG scaffolds for both cMSCs and hMSCs. However, the only difference between PEG and PEG-PDMS constructs that rose to statistical significance was for matrix mineralization (p = 0.002). Furthermore, Col-1 levels in the PEG-PDMS constructs appeared to be significantly lower than in corresponding PEG controls for both hMSCs and cMSCs, although these differences were only significant for hMSCs (p = 0.002).

Figure 2.

Osteogenic marker levels associated with 21 days of (A) human MSC or (B) canine MSC culture within PEG or PEG-PDMS scaffolds. All data are normalized to the PEG day 21 group. ‘$’ denotes a significant difference relative to the “day 0” control. ‘*’ denotes a significant difference relative to the PEG construct group, p < 0.05.

3.3 PEG-PDMS-BMP2 scaffolds for osteoblastic differentiation of canine MSCs

As a second goal of the current work, validated cMSCs were utilized to assess the efficacy of tethered, low-dose BMP2 in enhancing the intrinsic osteogenic capacity of PEG-PDMS scaffolds. A low dose of BMP2 (100 ng/mL) was selected for incorporation: 1) to reduce the potential for the undesired side effects currently associated with high dose BMP2 loading ⁷⁶, and 2) based on literature indicating 100 ng/mL doses could be effective in stimulating Col-1 deposition as well as calcification ^77,78.

To assess possible synergistic effects between PDMS and BMP2, the responses of cMSCs to three construct groups were compared: 1) PEG-PDMS, 2) PEG-BMP2, and 3) PEG-PDMS-BMP2. As with the incorporation of PDMS, the addition of BMP2 into the PEG network structure did not substantially impact scaffold bulk modulus or average mesh size (Table 1), again allowing focus on added bioactivity in comparing cell responses. As anticipated, cMSCs in PEG-BMP2 scaffolds displayed significantly increased Col-1 deposition (p < 0.001) relative to PEG-PDMS constructs (Figure 3). However, PEG-BMP2 networks did not stimulate statistically significant increases in runx2 or OPN relative to PEG-PDMS scaffolds. Furthermore, although the levels of the mid-term osteogenic transcription factor osterix appeared to be increased in PEG-BMP2 constructs relative to PEG-PDMS scaffolds, this difference fell below statistical significance (p = 0.269).

Figure 3.

Comparison of cMSC osteogenic responses following 21 days of culture in PEG-PDMS, PEG-BMP2, or PEG-PDMS-BMP2 scaffolds. All data are normalized to the PEG-PDMS group. ‘$’ denotes a significant difference relative to PEG-PDMS controls. ‘*’ denotes a significant difference relative to the PEG-BMP2 group.

When PDMS and BMP2 were combined, however, osterix levels increased 2.2-fold (p = 0.008) and calcium deposition increased 1.9-fold (p = 0.038) relative to pure PEG-PDMS constructs. PEG-PDMS-BMP2 scaffolds also displayed a 2.0-fold increase in OPN deposition (p < 0.009) relative to both PEG-PDMS and PEG-BMP2 constructs. That said, the addition of BMP2 to PEG-PDMS networks did not promote a significant increase in Col-1 relative to PEG-PDMS only constructs, despite the increase in Col-1 noted with BMP2 alone. The cumulative results from the first (Figure 2) and second (Figure 3) experimental sets indicate that PDMS may have a suppressive effect on Col-1 production. Representative images of Col-1, OPN, and von Kossa staining are presented in Supplementary Figure 1.

To assess the specificity of cMSC differentiation towards the osteoblastic lineage, markers for adipogenesis (AFABP) and chondrogenesis (Col-2, Col-10) were also assessed (Figure 4). Although no decrease in AFABP expression was observed in PEG-BMP2 scaffolds relative to PEG-PDMS controls (p = 0.089), PEG-PDMS-BMP2 scaffolds were associated with a significant reduction in AFABP expression relative to PEG-PDMS constructs (p = 0.001). No significant differences in Col-2 expression were noted for PEG-BMP2 or PEG-PDMS-BMP2 scaffolds relative to PEG-PDMS constructs. However, just as PDMS appeared to have a suppressive effect on Col-1 production, PDMS presence in the PEG-PDMS-BMP2 constructs was correlated with reduced levels of the hypertropic chondrocyte marker Col-10 relative to PEG-BMP2 scaffolds (p = 0.006). BMP2 is known to stimulate chondrogenic as well as osteogenic differentiation, with increases in the chondrogenic transcription factor sox9, the cartilage ECM protein Col-2, and the hypertropic chondrocytic marker Col-10 frequently being noted following low-dose BMP2 stimulation ^78–81. Thus, the present AFABP and Col-10 data indicate that PDMS incorporation may improve the osteogenic specificity of BMP2 bioactivity. Representative images of Col-2 immunostaining are presented in Supplementary Figure 1.

Figure 4.

Comparison of cMSC chondrogenic or adipogenic responses following 21 days of culture in PEG-PDMS, PEG-BMP2, and PEG-PDMS-BMP2 scaffolds. Levels of Col-2, Col-10 and AFABP relative to PDMS hydrogels. All data are normalized to the PEG-PDMS group. ‘$’ denotes a significant difference relative to PEG-PDMS controls. ‘*’ denotes a significant difference relative to the PEG-BMP2 group.

Although data from a single canine donor are shown in Figures 3 & 4, confirmatory experiments with PEG-PDMS-BMP2 scaffolds were conducted for an additional two cMSC donors. A high degree of correlation in type and degree of response with respect to scaffold formulation was observed among donors for each of the markers assessed (runx2, Col-10, and AFABP; Supplementary Figure 2).

4. DISCUSSION

The success of MSC-based scaffolds for bone regeneration largely depends on the associated MSCs supporting new bone formation. Successful clinical translation of MSC-based bone grafts based on in vivo preclinical data requires not only a similarly in anatomy and healing responses between humans and the selected animal model, but also a similarity in their MSC responses to scaffold-based stimuli. However, little is known about the degree of correlation in osteogenic responses of human and canine MSCs to 3D scaffold environments. In the current study, human and canine MSCs were cultured in vitro and subjected to equivalent 3D scaffold environments for 21 days in the absence of osteogenic media supplements. Following culture, scaffolds were analyzed for protein levels of several osteogenic markers as well as for deposited calcium.

Experimental results showed a high degree of agreement in cMSC and hMSC osteogenic responses to RGDS-containing PEG and PEG-PDMS scaffolds, indicating that canine models may warrant further development as preclinical animal models for MSC-based bone graft evaluation. Specifically, the culture of cMSCs and hMSCs in PEG constructs for 21 days promoted accumulation of Col-1 and calcium deposition. Furthermore, the addition of PDMS to PEG-based constructs increased OPN production and matrix mineralization by both cMSCs and hMSCs relative to day 0. The current PEG scaffold results are consistent with previous reports demonstrating that PEG constructs containing tethered RGD were able to facilitate MSC osteogenic differentiation and matrix mineralization over time ^82,83. The PEG-PDMS results also support previous work demonstrating siloxane and/or PDMS to intrinsically support matrix mineralization ^39,84–86 and to have an overall osteoinductive effect on human, rabbit and murine mesenchymal progenitor cells ^41,85–87. That said, the present data indicated Col-1 deposition in the PEG-PDMS constructs to be lower than that in corresponding PEG controls for both cMSCs and hMSCs. This latter result is in contrast to previous work with PEG-PDMS scaffolds and murine mesenchymal progenitor cells in which a positive correlation between Col-1 levels and PDMS incorporation was observed ⁴¹. This distinction underscores the need for appropriate animal model selection for both in vitro and in vivo testing.

In order to harness the positive effects of PDMS on calcium and OPN deposition while potentially minimizing its negative impact on Col-1 levels, low-dose BMP2 was incorporated into the PEG-PDMS constructs. Specifically, BMP2 was incorporated at 100 ng/mL, a dose ∼10,000-fold lower than that associated with the BMP2-loaded collagen sponges currently approved for certain in vivo applications ⁵³. The incorporated BMP2 was covalently-conjugated into the scaffold network in order to take advantage of the prolonged BMP2 retention and the potential for repeated signaling that tethering introduces ³¹. PEG-PDMS-BMP2 scaffolds were associated with a 2.2-fold increase in osterix expression, a 2.0-fold increase in OPN production, and a 1.9-fold increase in calcium deposition relative to pure PEG-PDMS systems. These results (obtained in the absence of osteogenic media supplements) are consistent in degree with the osteogenic effects previously observed following incorporation of substantially higher doses of tethered BMP2 (200–500 ng/mL) within various scaffold systems (usually cultured in the presence of osteogenic media supplements) ^{31,50,59,88,89}. For instance, Liu et. al. ³¹ reported that tethering 500 ng/ml of BMP2 into PEG-based scaffolds resulted in a ∼1.5-fold increase in calcium deposition after 12 days of culture in osteogenic induction medium. We report a 1.9-fold increase with 100 ng/mL BMP2 addition to PEG-PDMS constructs cultured for 21 days in the absence of osteogenic supplements. By way of further comparison, addition of 0.1 mg/ml of BMP2 (1000 times the current dose) to thiolated chitosan scaffolds promoted a 4-fold increase in calcium deposition after 21 days of culture in osteogenic media ⁹⁰.

In contrast to matrix mineralization, the addition of BMP2 to PEG-PDMS networks did not promote a significant increase in the osteogenic ECM protein Col-1 relative to PEG-PDMS scaffolds, despite the marked increase in Col-1 noted with the incorporation of BMP2 alone. In addition, the addition of PDMS to pure PEG scaffolds appeared to reduce Col-1 deposition relative to PEG only controls. These results indicate that PDMS may have a suppressive effect on Col-1 production. Similarly, assessment of the hypertrophic chondrocytic marker Col-10 in the PEG-PDMS-BMP2 scaffolds revealed that the presence of PDMS appeared to limit the production of Col-10 stimulated by BMP2 alone. Cumulatively, the present results indicate that PEG-PDMS-BMP2 constructs successfully promote specific osteogenic differentiation of cMSCs even in the absence of osteogenic media supplements.

In terms of mechanism, a combination of substrate-based stimuli along with growth factor signaling could potentially underlie the enhanced MSC osteogenic response noted for the PEG-PDMS-BMP2 scaffolds, although the present work cannot definitively state the relative contributions of each. In terms of substrate-based stimuli, Si-based additives have generally been demonstrated to be osteoinductive through increasing HAp deposits ^{39,85,86,91,92} as well as through altering the local nanotopography and hydrophobicity experienced by MSCs ⁹³. In terms of BMP2 signaling, BMPs exert their effects through binding to cell surface receptors and forming a heterotetrameric complex comprised of two dimers of type I and type II receptors. The formation of this complex initiates a signal transduction cascade which results in functional changes within the cell without the need for BMP internalization or translocation to the nucleus ^94,95. Further supporting this mechanism of signal transduction, Liu et al. demonstrated enhanced osteogenic differentiation (osteocalcin, OPN, and BMP2 expression) in MSCs treated with tethered-BMP2 relative to MSCs treated with equivalent levels of soluble BMP2 ³¹. This enhanced bioactivity of conjugated BMP2 relative to soluble BMP2 is believed to result from the prolonged signaling enabled by tethering, which hinders of BMP2 uptake and/or degradation ³¹. Given the similar bulk tensile moduli and average mesh size across scaffold formulations, the beneficial MSC response noted in the PDMS-BMP2 scaffolds likely reflects the combined effects of cellular interactions with tethered BMP2, the induced HAp deposits, and the material nanostructure. However, addition studies would be needed to confirm these potential mechanisms of action. Future work utilizing nanoscale mechanical testing with atomic force microscopy could help decouple the contributions from these factors.

A few limitations to the present study merit comment. First, the current work focused solely on hMSCs and cMSCs and did not investigate potential species differences among supporting cell types (e.g. osteoblasts osteoclasts, osteocytes, and macrophages) that are also important for regeneration. It should be noted, however, that although some species differences have been found for human versus canine MSCs ^38,96 and macrophages ⁹⁷, the overall degree of similarity between bone composition ⁹⁸, cell and tissue response to stimuli ^99–107, and pathological outcomes ^108–111 between these two species highlights a significant overlap in behaviors. Furthermore, evaluation of additional hMSC and cMSC donors may be needed for definitive conclusions to be made regarding inter-species similarity in the osteoinductive effects of PEG-PDMS scaffolds. This is due to conflicting literature regarding age and gender effects on MSC osteogenic responses, with some studies indicating significant differences with age and gender ¹¹² and others reporting a lack of significant differences in age and gender ¹¹³. As such, future work will confirm hMSC and cMSC responses to PEG-PDMS and PEG-PDMS-BMP2 scaffolds utilizing additional donors to more fully control for both age and gender.

5. CONCLUSIONS

The current study investigated, for the first time, the correlation between the osteogenic responses of cMSCs and hMSCs cultured within PEG scaffolds containing osteoinductive PDMS. The similarity in the osteogenic responses observed between cMSCs and hMSCs provides strong initial support for the use of cMSCs for the in vitro evaluation of MSC-based bone grafts. Using cMSCs, we further evaluated the efficacy of low-dose, tethered BMP2 in enhancing the osteoinductivity of PEG-PDMS scaffolds. The combined presence of PDMS and BMP2 significantly increased osterix, OPN, and mineralization levels relative to PEG-PDMS scaffolds while reducing the chondrogenic effects of BMP2 alone. These data suggest that low-dose BMP2 can be effective in driving osteogenesis when coupled with an intrinsically osteoinductive material.