Macropore Design of Tissue Engineering Scaffolds Regulates Mesenchymal Stem Cell Differentiation Fate

Abstract

Craniosynostosis is a debilitating birth defect characterized by the premature fusion of cranial bones resulting from premature loss of stem cells located in suture tissue between growing bones. Mesenchymal stromal cells in long bone and the cranial suture are known to be multipotent cell sources in the appendicular skeleton and cranium, respectively. We are developing biomaterial constructs to maintain stemness of the cranial suture cell population towards an ultimate goal of diminishing craniosynostosis patient morbidity. Recent evidence suggests that physical features of synthetic tissue engineering scaffolds modulate cell and tissue fate. In this study, macroporous tissue engineering scaffolds with well-controlled spherical pores were fabricated by a sugar porogen template method. Cell-scaffold constructs were implanted subcutaneously in mice for up to eight weeks then assayed for mineralization, vascularization, extracellular matrix composition, and gene expression. Pore size differentially regulates cell fate, where sufficiently large pores provide an osteogenic niche adequate for bone formation, while sufficiently small pores (<125 um in diameter) maintain stemness and prevent differentiation. Cell-scaffold constructs cultured in vitro followed the same pore size-controlled differentiation fate. We therefore attribute the differential cell and tissue fate to scaffold pore geometry. Scaffold pore size regulates mesenchymal cell fate, providing a novel design motif to control tissue regenerative processes and develop mesenchymal stem cell niches in vivo and in vitro through biophysical features.

Graphical Abstract

Introduction

Craniosynostosis is a debilitating disease, characterized by the premature fusion of growing cranial bones, affecting 1 in 2,500 live births. Craniofacial defects are the most common congenital birth defects in the U.S. – posing significant costs to patients and their families [1, 2]. Premature bone fusion makes the skull unable to accommodate cranial vault expansion necessary for the growing brain. Therefore, craniosynostosis can result in serious effects on infant and childhood development including: high intracranial pressure, blindness, and respiratory impairment, making the disease important to recognize and treat [2–4]. Currently, treatment requires invasive cranial surgeries leading to high morbidity, particularly for patients requiring multiple surgeries where a single surgery is unable to fully correct abnormalities [2, 5–7]. Despite advancements in cranial suture biology, no pharmacologic treatments are available.

The cranial suture is a fibrous tissue that connects bones of the skull (Scheme 1), composed of neural crest and mesoderm derived suture mesenchymal stem cells (SMSC, red in Scheme 1) [8]. With improved understanding of cranial suture biology and craniosynostosis etiology, clinicians and patients are eager for advanced therapeutics with the potential to robustly treat and/or diminish severity of this debilitating disease. Maintenance of stem cells (MSCs) in the cranial mesenchymal suture tissue that exists between growing cranial bones is critical for cranial bone growth [9–12]. Recent literature points a possibility of craniosynostosis as a stem cell disease, where disease progression is caused by aberrant and dysregulated cell fate specification and differentiation of the SMSC population [10, 13, 14]. Compared to wild-type mice, we and others demonstrated in several craniosynostosis mouse models that sutures which fuse early in mouse development are less organized and vascularized [13, 15]. Zhao et al. demonstrated that SMSCs behave as MSCs in vitro and characteristically express Gli1, able to differentiate towards osteogenic, chondrogenic, and adipogenic fates, and are capable of self-renewal to repair damaged cranial bone. Ablation of the Gli1+ SMSC population gives way to craniosynostosis and disrupted craniofacial morphogenesis [16]. Therefore Gli1+ SMSC residing in the cranial suture are hypothesized to be a unique skeletal stem cell population, with similar function to recently identified skeletal stem cells in long bone [17] [18]. Based on these observations and previous work by our group [19], we hypothesized that a tissue engineering strategy to sufficiently maintain the Gli1+ SMSC population in vivo, necessary to maintain the patent cranial suture and to function in cranial osseous wound healing, would require distinct three-dimensional microenvironments based on scaffold pore design. In a tissue engineering approach to treating craniosynostosis, the ideal biomaterial construct would present a microenvironment which predictably preserves the cranial suture mesenchyme as a distinct stem cell population, in addition to providing a surrounding microenvironment which facilitates osteogenesis and bone growth. These facts indicate a necessity of a specialized scaffold design to provide a distinct and contrasting microenvironment from those designed to facilitate osteogenic differentiation and bone repair. A cell-scaffold construct which recapitulates the cranial suture mesenchyme represents an opportunity to significantly improve long term quality of life for craniosynostosis patients.

Scheme 1.

The cranial suture mesenchyme is a connective fibrous tissue of the skull and cranial stem cell niche, flanked by cranial bone. The cranial suture mesenchyme is made up of suture mesenchymal stem cells, postulated to be the skeletal stem cell niche of the calvaria. Flanking osteogenic fronts are lined with mineralizing osteoblasts and pre-osteoblasts, with mature osteocytes embedded in the bone matrix.

Tissue engineering scaffolds serve as a temporary, artificial extracellular matrix (ECM) which provides a physical and biological microenvironment for cell-based therapies. Nanofibrous, macroporous tissue engineering scaffolds are fabricated from poly (L-lactic acid) (PLLA) [20]. Nanofibers are the result of thermally induced phase separation (TIPS) of PLLA from organic solvent at low temperatures, resulting in fibrillar morphology with fiber diameters on average 50–500 nm, analogous to the native collagen ECM [21]. Tissue engineering scaffold properties are critical engineering considerations for modulating regenerative outcomes. Biomaterial stiffness [22, 23], void space and porosity [24], and pore size [25–29] are recognized parameters that affect cell fate and desired tissue outcomes. Relevant to this study, interconnected macropores enabled by a sugar sphere porogen method allows for well-controlled design of pore shape and size to investigate ideal microenvironments which influence cell and tissue fate [19, 28]. The interconnected pore structure allows for robust three-dimensional tissue formation, uniform cell seeding, and nutrient/waste transfer, enabling efficient tissue regeneration.

Previously we demonstrated that macropore size is a critical parameter which modulates chondrogenesis and vascularization, where pores of <125 um prevent blood vessel penetration and maintain a cartilage phenotype after sufficient in vitro induction, while pores of > 250 um facilitate vascularization and endochondral ossification of the engineered tissue construct to form robust bone from a cartilage template [19]. To date, creation of an optimized tissue engineering scaffold to recapitulate and regenerate an intramembranous bone stem cell niche is an underdeveloped area of regenerative medicine [30]. Maintenance of a cranial bone stem cell phenotype, that of the SMSC, is an exciting tissue engineering opportunity for translation to future in vivo applications to support cranial bone and suture regeneration, among other skeletal applications such as facial bones.

Herein we aim to develop and characterize three-dimensional scaffold-based microenvironments which appropriately recapitulate the cranial suture mesenchyme and surrounding bone phenotypes, separately, towards development of a tissue engineering-based treatment for craniosynostosis. Our central hypothesis is that scaffold pore size influences the regenerative fate of the cell-scaffold microenvironment towards stemness or differentiation, by modulating cell and tissue phenotype. We report that pores less than 125 um in diameter prevent premature differentiation of MSCs by preventing vasculature ingrowth and favoring an immature matrix, without in vitro induction. In contrary, sufficiently large pores, greater than 250 um in diameter, support a trajectory towards an osteogenic niche and mineralized bone formation by encouraging differentiation, secretion of a mature extracellular matrix, and robust vascularization. This data indicates that provision of scaffolds containing specified regions of different pore sizes is a useful tissue engineering strategy to maintain stemness of cranial suture cells while also promoting cranial bone growth. Using such a tailored biomaterial platform, we are working towards provision of appropriate microenvironmental cues to both study and develop treatments for individuals with craniosynostosis.

Materials + Methods

Materials:

Poly (L-lactic acid) (PLLA, Resomer L207S) with an inherent viscosity of 1.6 dl/g was purchased from Boehringer Ingelheim (Ingelheim, Germany), Span80, ascorbic acid, β-glycerophosphate, human recombinant insulin, sodium phosphate, isobutylmethylxanthine, dexamethasone, troglitazone, Alizarin Red-S, Oil Red-O solution, Alcian Blue, acetic acid, hydrochloric acid, dimethyloxalylglycine (Sigma); tetrahydrofuran (THF), hexane (Fisher Scientific); D-fructose (Oakwood Chemical); mineral oil (Alfa Aesar). All reagents were used as received unless otherwise noted.

Fabrication of Three-Dimensional Macroporous Scaffolds:

Nanofibrous, macroporous tissue engineering scaffolds were fabricated from poly (L-lactic acid) (PLLA) as previously described [20]. Briefly, PLLA was dissolved in tetrahydrofuran solvent (THF, 10% wt/v, 60°C). Separately, D-Fructose is heated to melt and is emulsified in hot mineral oil with Span80 surfactant by mechanical stirring. The fructose-mineral oil mixture is quenched with ice to solidify the sugar spheres resulting from the emulsion. Sugar spheres were washed thoroughly with hexane to remove mineral oil and Span80 surfactant. Sugar spheres were separated by size with molecular sieves (Newark Wire Cloth Co – 63 um, 125 um, 250 um, 425 um) and loaded into Teflon molds. The sugar spheres in hexane were annealed at 37°C for a prescribed amount of time (7–11 minutes depending on sphere size) to cause the spheres to adhere to each other, introducing pore interconnectivity. Hexane was removed under vacuum and PLLA (10% wt/v in THF) was cast and immediately chilled to −80°C for 48 hours to induce phase separation. After 48 hours, sugar-polymer constructs were transferred to hexane to exchange the THF solvent for an additional 24 hours, then removed from the Teflon vial and soaked in water for 24 hours, to completely remove the sugar porogen. The result is a nanofibrous, macroporous three-dimensional tissue scaffold which can be cut to size with a biopsy punch. The morphology of scaffolds was examined by scanning electron microscopy.

Scaffold Sterilization:

Prior to all in vitro and in vivo work, PLLA scaffolds were sterilized by a dual-sterilization method. First, constructs were sterilized by ethylene oxide gas according to the manufacturer’s protocol (Anpro). Secondly scaffolds were washed with 70% ethanol for 30 minutes, followed by washing with PBS then with cell culture media, immediately before cell seeding. The purpose of the ethanol wash is twofold – first, a secondary sterilization method, and second, to “wet” the surface of the hydrophobic PLLA scaffold prior to cell seeding.

Scanning Electron Microscopy (SEM):

SEM imaging was used to evaluate the nano- and micro-scale surface topographies of PLLA scaffolds. Scaffolds were affixed to sample holders with double-sided carbon tape, and gold coated (120 s, DeskII, Denton Vacuum) and observed at 5 kV with a working distance of 10 mm (JEOL JSM-7800FLV).

Determining Compressive Modulus of Scaffolds:

Mechanical properties of scaffolds were measured using an MTS Synergie 200 mechanical tester (MTS Systems, Inc.). Scaffolds with dimensions of 15.0 mm in diameter and 3.0 mm in thickness were prepared, n=5 per pore size. Compressive modulus was defined as the initial linear modulus on the resulting stress-strain curve, with a strain rate of 1.0 mm/minute.

Isolation of Primary Cells and Culture Conditions:

Primary suture mesenchyme stem cells (SMSCs) were isolated from sagittal sutures containing its adjacent parietal bones of 3- to 5-day-old mice. In brief, suture tissues were incubated with 0.2% collagenase in PBS at 37 °C for 1 h. The dissociated cells were filtered through a 40 μm strainer and cultured in alpha-minimum essential medium (aMEM) supplemented with 10% fetal bovine serum (Denville Scientific) and 1% penicillin/streptomycin (Invitrogen). Primary bone marrow stromal cells (BMSCs) were isolated from femora and tibiae of 3-week-old mice and then cultured in Dulbecco’s Modified Eagle Medium (DMEM) containing 10% FBS. All cells were used for experiments before passage 4. Their character as MSCs was characterized as described below as well as gene expression profile, shown in Fig 2, respectively. These procedures were performed following a protocol approved by the University of Michigan Institutional Animal Care and Use Committee (IACUC).

Figure 2. MSC characterization.

The mesenchymal stem cell identity and functional phenotype of both BMSCs, skeletal MSCs of long bone, and SMSCs, skeletal MSCs of the calvaria, is characterized. Primary BMSC and primary SMSC populations both are highly enriched for CD45-;CD44+;CD29+ cells, determined by flow cytometry, conforming to characteristic MSC identity (A). Both BMSCs and SMSCs demonstrate trilineage differentiation potential to osteogenic, chondrogenic, and adipogenic cell fates following in vitro induction for 14 days. Osteogenic fate is determined by Alizarin Red staining, chondrogenic fate is determined by Alcian Blue staining, and adipogenic fate is determined by Oil Red staining as indicated by arrows pointing to lipid droplets (B). Scale = 200 um. SMSC and cranial osteoblasts were cultured (separately) in growth and osteogenic cell culture media, in 2D tissue culture dishes, to ensure their characteristic identities and create a baseline for gene expression analyses in cell-scaffold constructs (2C, n=4). Values represent the mean ± SD. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05.

Fluorescence-Activated Cell Sorting (FACS):

For analysis of mesenchymal stem cell (MSC) phenotype [31], BMSC and SMSCs were assayed for cell surface markers: CD45, CD44, CD29. 1 × 10⁶ BMSC and SMSC were seeded onto a 10 cm cell culture plastic dish and cultured for 5 days in growth media. Cells were incubated with APC/Cy7 anti-mouse CD45 antibody (cat. 103116, Biolegend, San Diego, CA), APC anti-mouse CD44 antibody (cat. 103012, Biolegend), FITC anti-mouse CD29 antibody (cat. 102206, Biolegend) for 30 min at 4C. 4’,6-diamidino-2-phenylindole (DAPI, cat. NBP2–31156, NOVUSBIO, Centennial, CO) was used for determining cell viability. The cell phenotypes were evaluated by FACS analysis for DAPI−;CD45−;CD44+;CD29+ (FACSAria Ilu Flow Cytometer, Becton Dickinson, Mountain View, CA).

Trilineage Differentiation Assay:

BMSC and SMSC were cultured on tissue plastic to reach confluence, in growth media. Once cells reached confluence, they were treated with either osteogenic, chondrogenic, or adipogenic differentiation media for 14 days. Their phenotype was evaluated by characteristic in vitro staining as described (n=3 per cell type, per treatment group). Osteogenic Differentiation: Once cells reached confluence, growth media was exchanged for osteogenic media (α-MEM, 10% fetal bovine serum, 1% penicillin/streptomycin supplemented with 50 ug/mL ascorbic acid and 10 mM β-glycerophosphate). Media was changed every 2 days for 14 days, until mineral nodule formation was observed. Chondrogenic differentiation: Once cells reached confluence, growth media was exchanged for chondrogenic media (DMEM/F12, 10% fetal bovine serum, 1% penicillin/streptomycin supplemented with 50 ug/mL ascorbic acid, 100 ug/mL insulin, and 5 mM sodium phosphate). Media was changed every 2 days for 14 days. Adipogenic differentiation: Once cells reached confluence, growth media was exchanged for adipogenic induction media (DMEM high glucose, 10% fetal bovine serum, 1% penicillin/streptomycin, supplemented with 0.5 mM isobutylmethylxanthine prepared from DMSO stock solution, 1 uM dexamethasone prepared from ethanol stock solution, 100 ug/mL insulin, and 10 uM troglitazone prepared from DMSO stock solution). After 3 days of culture, induction media was exchanged for adipogenic maintenance media (DMEM high glucose, 10% fetal bovine serum, 1% penicillin/streptomycin supplemented with 100 ug/mL insulin). This cycle of induction-maintenance was repeated twice, until round adipocyte morphology was observed.

Alizarin Red Assay, In Vitro:

After 14 days of osteogenic culture, media was removed and the cells were washed with phosphate buffer solution (PBS, pH 7.4) three times, then fixed with 10% formalin for 10 minutes. Fixed cells were washed three times with ultrapure water and incubated for 15 minutes with Alizarin Red-S solution (20 g/L, pH 4.2). After incubation, stained cells were washed three times with PBS to remove excess staining solution, and imaged with a bright field microscope and digital camera.

Alcian Blue Staining, In Vitro:

After 14 days of chondrogenic culture, media was removed and the cells were washed with phosphate buffer solution (PBS, pH 7.4) three times, then fixed with 10% formalin for 10 minutes. Fixed cells were washed three times with ultrapure water and incubated with Alcian Blue solution (0.1 g in 10 mL 3% acetic acid) for 60 minutes. After incubation, the stained cells were washed with 1 M HCl for 3 minutes, then washed thoroughly with PBS before imaging with a bright field microscope and digital camera.

Oil Red-O Staining, In Vitro:

After two cycles of adipogenic induction-maintenance culture, media was removed and the cells were washed with PBS three times, then fixed with 10% formalin for 10 minutes. Fixed cells were washed three times with ultrapure water, washed with a 60% isopropanol solution for 5 minutes, then incubated with Oil Red-O staining solution (0.5 g/100 mL isopropanol) for 15 minutes. Excess stain was removed and the cells were washed with PBS three times. Cells were left covered in PBS and imaged with a bright field microscope and digital camera.

Cell Culture of 3D Tissue Constructs:

Scaffolds were soaked in 70% ethanol for 30 min and washed three times with PBS for 10 min each, then washed with cell culture media. Primary cells (2.0 × 10⁵ cells/scaffold) were seeded onto each side of the nanofibrous scaffold (5 mm diameter 1.5 mm height). One hour later, culture media was gently added to cover the 3D constructs. Constructs were maintained for 24 hours and then incubated with culture media containing 10% (v/v) alamarBlue reagent (Invitrogen) for 4 hours. At the end of cultivation, 100 μL of culture media was transferred to a 96-well plate in triplicate for each sample and levels of 580 nm emission were measured with 530 nm excitation using a fluorescence plate reader to quantify conversion of resazurin to resorufin as a surrogate of attached live cells on the constructs to establish cell seeding efficiency.

Subcutaneous Implantation in Mice:

250,000 primary cells were evenly seeded onto each nanofibrous scaffold (8 mm diameter × 1.5 mm height) as described above. Constructs were cultured in media for 24 hours then implanted subcutaneously into wild type male mice aged 8–10 weeks old. All animal procedures were performed following a protocol approved by the University of Michigan Institutional Animal Care and Use Committee (IACUC). Mice were anesthetized via isoflurane inhalation and a midsagittal incision was made on the dorsa of each mouse. On each side of the midline, two subcutaneous pockets were made by blunt dissection such that four cell-scaffold constructs were implanted into each mouse in distinct regions. Incisions were closed with surgical staples and animals were given analgesic medication (carprofen) to manage pain. Mice were monitored closely and showed no adverse signs. At 4- and 8-weeks’ time following subcutaneous implantation, mice were sacrificed by inhalation of CO₂ and bilateral pneumothorax. Constructs were carefully explanted and either fixed in 4% paraformaldehyde prior to subsequent histologic processing or homogenized in TRIzol (Ambion) for RNA extraction and gene expression analysis. All animal procedures were prospectively approved of by the Institutional Animal Care and Use Committee (IACUC) at the University of Michigan.

Histology:

All samples were fixed in 4% paraformaldehyde to fix tissue, then transferred to 70% ethanol for dehydration before embedding in paraffin. Serial sections were cut at 5 um thickness. Standard protocols were followed for: hematoxylin and eosin, Masson’s trichrome, Alizarin Red, Picrosirius Red, Alcian Blue and Safranin O staining.

Immunohistochemistry:

Deparaffinized sections were placed in 10 mM citrate buffer (pH 6.0) for antigen retrieval. After treatment with 3% hydrogen peroxide and blocking solution, the sections were incubated with the primary antibody (CD31: Cell Signaling; Endomucin: Santa Cruz; CD146: Proteintech) overnight at 4°C. The sections were treated with HRP-conjugated goat anti-rabbit IgG (Abcam) according to the manufacturer’s instructions. The nuclei were stained with hematoxylin.

Image Analysis:

All image analysis was carried out in Fiji imaging software (Image J Image J, V 1.0.0-rc-69/1.52p). Images were imported as raw files (.TIF). Analyses were carried out using batch macros following optimized protocols; the contents of each automated macro are given in the supplementary information.

CD31+ Area by Immunohistochemistry:

Immunohistochemistry images were first acquired using a digital camera attached to a bright field microscope across all dimensions of the scaffold construct. TIF images were converted to an RGB stack. The stack was converted to images, and the blue channel was used for analysis. The total image was set to threshold 0–255 and measured, then was set to a new threshold to 0–125 which selectively identified CD31+ area, but not cell nuclei or background, and measured. The total CD31+ area was calculated as a ratio:

$$r_{CD31+}=\frac{are{a}_{t(0-125)}}{are{a}_{t(0-255)}}$$

Masks were created of all threshold areas and checked to ensure accuracy. Data are reported as fold change in r_CD31+ compared to small pore scaffolds at each time point. A minimum of n=20 images were analyzed in each group.

Blood Vessel Penetration Depth:

Immunohistochemistry images were taken around the border of paraffin sections. The scale of the image was calibrated to the magnification of the microscope image to convert pixel distances to microns. Distances were measured from the edge of the synthetic polymer construct in a straight line to the center of nearby blood vessels penetrating into the construct. A minimum of n=70 measurements were made across a minimum of n=10 images per group.

Blood Vessel Diameter:

High magnification immunohistochemistry images used for quantifying CD31+ area were also used to calculate blood vessel diameter. Raw images (.TIF) were scaled appropriately to convert pixel distances to microns, then the diameter of each blood vessel was measured as a straight line across the widest part. Only blood vessels which were round in shape (i.e. perpendicular to the section) were measured A minimum of n=90 measurements were made across a minimum of n=10 images per group.

Collagen Amount by Masson’s Trichrome Staining:

Masson’s Trichrome-stained histologic sections were imaged with an Olympus BX51 microscope and digital camera. Images were converted to 8-bit RGB images and histogram values were extracted for nBins=256, based on methods previously published [32]. Methyl blue is the active ingredient in Massons’ Trichrome Stain which positively stains collagen components of the extracellular matrix. Thresholding was optimized manually to select for Methyl Blue positive area and not to include Fast Red positive area or background signal. Area fraction was computed to determine the amount of Methyl Blue positive area per image, relative to the total sum of pixel intensities across all 256 bins, as a proxy for collagen amount per image:

$$r_{col+}=\frac{{\sum }_{i=n}^m{I}_i}{{\sum }_{j=0}^{256}{I}_j}$$

Extracellular Matrix Amount and Composition by Picrosirius Red:

Picrosirius red (PSR) stained images were imaged with an Olympus BX51 microscope and digital camera, equipped with a polarized light filter. For analysis, images were converted into an HSB stack and the Hue image was selected. The image was binned into 256 unique bins which represent pixel counts for each of 256 hues and plotted. The values of each bin were recorded. A minimum of n=10 images were analyzed in each group.

Four colors were considered, based on methods previously published [33]. Picrosirius red staining color reflects the size of collagen fibers, related to their crosslinking and maturity [34]. Red and orange color indicates mature collagen, while green and yellow color indicates immature matrix components [35]. Colors were binned as follows: Red = 2–9 and 230–256; Orange = 10–38; Yellow = 39–51; Green = 52–128. Hues from 129–229 were excluded and attributed to non-birefringent tissue elements. Relative area fraction by color was computed for each group to determine the relative abundance of mature to immature collagen in each histologic sample [35], relative to the total amount of PSR+ staining:

$$r_c=\frac{{\sum }_{i=n}^m{I}_i}{I_{PSR+}}$$

Where n and m are the lower and upper limits of the color bin. The total amount of extracellular matrix, based on PSR+ staining, was calculated as:

$$I_{PSR+}=\sum _{i=0}^{256}{I}_i-\sum _{j=129}^{229}{I}_j$$

Quantification of Alizarin Red Staining:

Alizarin Red stained slides were imaged with an Olympus BX51 microscope and digital camera. For analysis, images were converted into an HSB stack and the Hue image was selected. The image was binned into 256 unique bins which represent pixel counts for each of 256 hues and plotted. The values of each bin were recorded. A minimum of n=10 images were analyzed in each group. Red-positive area was selected for quantification of total AR+ area (230–256). Hues from 230–242 were considered “heavy” staining, and hues from 243–256 were considered “light” staining. Relative area fraction by color was computed for each group to determine the degree of mineralization:

$$r_c=\frac{{\sum }_{i=n}^m{I}_i}{I_{AR+}}$$

Where n and m are the lower and upper limits of the color bin.

Quantitative Gene Expression Analysis:

First-strand cDNA was synthesized from 500 ng of RNA using SuperScript II cDNA Synthesis Kits (Invitrogen) following the manufacturer’s instruction. Quantitative polymerase chain reaction (PCR) was performed using Power SYBR Green PCR Master Mix (Applied Biosystems). Levels of gene expression were compared among groups with Applied Biosystems ViiA7 platform. Expression levels of each gene were normalized to endogenous Gapdh. The primer sets are shown in Table S1. The amplification specificity was confirmed by melting curves. Trends in quantitative rt-PCR data was further analyzed using Morpheus (Broad Institute, https://software.broadinstitute.org/morpheus/) to generate heat maps of gene expression data. Similarity matrices were calculated by Pearson correlation for rows and columns of each heat map.

In Vitro Colorimetric Mineralization Assay:

Cell-scaffold constructs seeded with BMSC or SMSC were cultured for 3 weeks in vitro in osteogenic media. A cresolphthalein complexone colorimetric assay was used to determine the calcium content of constructs to assess mineralization, according to the manufacturer’s instructions (Pointe Scientific C7503). Briefly, cell-scaffold constructs were washed in PBS, then incubated in 0.5 N HCl overnight at 4°C. 295 uL of reagent and 5 uL acidified extract were mixed in a 96-well plate and read with a UV-Vis plate reader (Thermo). Samples (n=8 per cell-scaffold combination, 32 total) were assayed in triplicate. Calcium concentration was read by 570 nm absorption, and calibrated against a standard curve (CaCl₂).

Hypoxic Cell Culture Conditions in 2D Culture:

Cells were seeded and cultured to confluency on 10 cm tissue culture-treated plastic dishes. Once cells reached confluency, growth media was exchanged for osteogenic media and cells were cultured for 7 days. Culture media was changed every two days. After 7 days, cells were treated with various concentrations of dimethyloxalylglycine (DMOG) to mimic hypoxia, ranging from 0–2 mM.

Western Blot Analysis:

Whole cell lysates were separated on 4–20% Tris-Glycine polyacrylamide gel and transferred to PVDF membranes. The membranes were incubated with 5% bovine milk for 1 hour and incubated with primary antibodies overnight at 4°C (anti-HIF-1a cat. 3716, Cell Signaling, Danvers, MA; anti-VEGFB, cat. 2463, Cell Signaling). Blots were incubated with peroxidase-coupled anti-rabbit IgG secondary antibody (cat. 7074, Cell Signaling) for 1 hour, and protein expression was detected with SuperSignal West Dura Chemiluminescent Substrate (cat. Prod 34075, Thermo Scientific, Rockford, IL). Membranes were restained with monoclonal anti-β-actin antibody (cat. 4970, Cell Signaling) to control for equal loading.

Hypoxic Cell Culture Conditions in 3D Scaffolds:

Cell-scaffold constructs seeded with BMSC or SMSC were cultured in media containing 1mM DMOG for 3 weeks. Control constructs were cultured with media containing 0.1 % dimethyl sulfoxide (DMSO). RNA was extracted from the constructs and cDNA was synthesized using TRIzol and SuperScript II cDNA Synthesis. The primer sets are shown in Table S1.

Statistical Analysis:

All data are reported as mean ± standard deviation and represent a minimum sample size of n≤4. Statistical analysis was carried out in GraphPad Prism v8. Student’s t-test was used to determine statistical significance of observed values between experimental groups where p < 0.05 was considered significant. Tukey’s test was used to determine differences between group means as a single-step method to compare multiple means and determine statistical significance between. Statistical analyses were carried out under the guidance of the University of Michigan Consulting for Statistics, Computational and Analytical Research Center. In all graphics, significance is noted as: * p<0.05, ** p<0.01, ***p<0.001, ****p<0.0001.

Results

Nanofibrous, Macroporous Tissue Engineering Scaffolds of Discrete Pore Sizes

Representative scanning electron microscopic (SEM) images of highly porous three-dimensional tissue engineered scaffolds are shown in Figure 1. Scaffolds consist of uniformly spherical interconnected macropores, the result of sugar sphere porogen method, which allows for discrete control of spherical pore diameter (Fig 1A–C). We fabricated scaffolds of pore sizes: small (63–125 um). medium (125–250 um), and large (250–425 um). Interconnections between macropores allow for mass transport and cell migration throughout the constructs. The fabrication protocol was optimized to ensure consistent scaffold properties among different pore sizes, namely interconnectivity, to allow for cell migration and efficient mass transfer (Fig S1). By analysis of SEM images, we demonstrated that the relative amount missing surface area due to the interpore connectivity is consistent across pore sizes (Fig 1D–F, Fig S1A). At high magnification, scaffolds have a consistent nanofibrous surface, with fiber diameters between 50–500 nm, resulting from thermally induced phase separation (TIPS) of crystalline PLLA, previously shown to facilitate cell attachment, proliferation, and protein secretion, advantageous for tissue engineering applications (Fig 1G–I) [21]. The compressive modulus of scaffolds across different pore sizes is not statistically different (Fig S1B). All scaffold varieties are sufficiently stiff so that differential cellular response is attributed to physical morphology of constructs [36]. Regardless of pore size, scaffolds have a similar porosity of about 98%, resulting from the highly porous nanofibers [20].

Figure 1. Macroporous, nanofibrous scaffolds of controlled pore sizes from PLLA.

Morphology of three-dimensional tissue engineering scaffolds fabricated from poly (L-lactic acid) (PLLA) is observed by scanning electron microscopy (SEM). Scaffolds are macroporous, the result of sugar porogen method during fabrication, and macropore size is well-controlled based on selecting discrete subpopulations of sugar spheres by molecular sieves (A, B, C; Scale = 100 um). Macropores are interconnected, the result of a heat treatment method, allowing for nutrient/waste exchange and cell migration throughout the construct (D, E, F; Scale = 20 um). The surfaces of the scaffold are nanofibrous, the result of thermally induced phase separation of PLLA (G, H, I; Scale = 10 um).

Subcutaneous Implantation of BMSC and SMSC Cultured on Various Pore Size Constructs

Primary bone marrow stromal cells (BMSC) and suture mesenchyme stem cells (SMSC) were isolated from wild type mice, and characterized by flow cytometry and trilineage differentiation assay to confirm their phenotype (Figure 2). Flow cytometry analysis of early passage BMSC and SMSCs showed characteristic enrichment for CD45-;CD44+;CD29+ cells (Fig 2A). DAPI-negative cells (viable cells) were sorted for CD45-(47.5±0.9% BMSCs; 85.4±1.2% SMSCs). This population was further characterized as highly expressing CD44 (96.5±2.3% BMSCs, 99.3±0.1% SMSCs) and CD29 (98.6±0.6 BMSCs, 99.7±0.1% SMSCs). Cells were cultured in osteogenic, chondrogenic, and adipogenic induction media for 14 days (Fig 2B). Alizarin Red-positive calcium nodule formation was observed in cells subjected to osteogenic differentiation media for both cell types. Alcian Blue-positive sulfated glycosaminoglycan matrix was observed to be secreted by cells subjected to chondrogenic differentiation media, similarly for both cell types. Finally, Oil Red-positive lipid droplets were observed with typical large, round adipocyte morphologies after two cycles of adipogenic induction over the course of 14 days. Cells cultured in growth media did not show significant differentiation, for both cell types. Specific to osteogenic differentiation, both SMSC and BMSC show marked changes in gene expression of characteristic genes related to stemness and osteogenesis (Figure 2C).

Primary BMSCs or SMSCs were seeded onto scaffolds of three distinct pore sizes: small, medium, and large. After 24 hours of in vitro culture, cell seeding efficiency is comparable among different scaffolds for both cell types and sufficient time for cells to attach to the nanofibrous surface as demonstrated by scanning electron microscopy (Fig S1C–H, Fig S1I). After 24 hours of in vitro culture, cell-scaffold constructs were implanted subcutaneously in mice and kept for 4 and 8 weeks before explanting for histologic evaluation.

At 4 weeks, for both SMSC and BMSC, hematoxylin and eosin (H&E) staining demonstrate that cells are well-spread throughout the scaffold constructs, regardless of pore sizes (Fig 3A–B”). Cells are well-distributed throughout the porous spaces of the constructs, adhering tightly to the pore walls. Masson’s trichrome shows increased collagen extracellular matrix deposition in large and medium pore constructs compared to small pore constructs, as indicated by increased collagen (blue) staining, although these differences are insignificant by 8 weeks time (Fig 3C–D, I). BMSC secrete greater amounts of collagen than SMSC, particularly at earlier time points. Additionally, BMSC and SMSC show different optimal microenvironment for maximum collagen secretion in their response to microenvironment pore size (Fig 3I).

Figure 3. Histology of subcutaneous implanted constructs.

Macroporous, nanofibrous scaffolds of specified pore sizes were seeded with suture mesenchyme stem cells (SMSC) or bone marrow stromal cells (BMSC), then implanted subcutaneously in mice after 24 hours of in vitro induction to allow for cell adhesion. At 4 and 8 weeks, constructs were harvested and analyzed histologically by hematoxylin and eosin staining (A, B, E, F) and Masson’s trichrome staining (C, D, G, H). Scaffolds are well-cellularized across all combinations of pore size and cell type; extracellular matrix composition differs as a function of pore size based on methyl blue intensity in Masson’s trichrome staining. Scale = 200 um. Quantification of methyl blue intensity in Mason’s trichrome staining (I) where images (n > 10 per cell type and pore size) were analyzed by hue and thresholded to specifically select for colors associated with Methyl Blue, indicating collagen staining. Results are in agreement with total PSR+ area quantified in Figure 6, with slight variations in total extracellular matrix secreted as a function of pore size and cell type. Values represent the mean ± SD. At 4 weeks, at each pore size, ECM deposition is significantly different (**) by cell type, but insignificant at 8 weeks. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05.

H&E staining of constructs after 4 and 8 weeks of implantation demonstrates that both SMSC and BMSC continue to be uniformly spread throughout the constructs with an increasing cell density over time (Fig 3E–F). Cells in large pore diameter constructs seem to adopt a more spread, lamellar morphology along pore walls compared to cells in constructs with pore diameters less than 125 um. Masson’s trichrome staining at 8 weeks indicates consistent deposition of a rich collagenous extracellular matrix which supports future tissue formation (Fig 3G–H, I).

Scaffold Pore Size Regulates Construct Mineralization

Subcutaneous explants were stained with Alizarin Red (AR) at 8 weeks to assess the degree to which scaffold pore size facilitated or limited the mineralization potential of implanted cells (Fig 4). Based on threshold image analysis for AR-positive area, BMSC in 250–425 um diameter pore constructs mineralize their matrix more significantly than any other cell-pore combination. BMSC in constructs with pore size less than 250 um mineralized to a greater degree than SMSC in the same constructs, but both were significantly limited compared to constructs with pore size greater than 250 um. By image analysis we applied a threshold to segregate heavy and light staining of Alizarin Red positive area. Heavy (most intense) staining is indicative of intense mineralization by cell clusters, critical for robust bone formation. Light staining indicates disperse matrix calcification. This data demonstrates that pores less than 125 um in diameter prevent functional differentiation of an osteogenic tissue phenotype in vivo for both cell types. Small pores inhibit matrix mineralization, while sufficiently large (>250 um) pores facilitate robust mineralization. The same constructs were subjected to both Safranin O and Alcian Blue staining to evaluate for a potential chondrogenic phenotype (Fig S2). Interestingly, there is no apparent difference in chondrogenic phenotype as a function of pore size, for either cell type, after 8 weeks of in vivo subcutaneous culture.

Figure 4. Alizarin Red Histology + Quantification.

Histologic sections from subcutaneously implanted constructs at 8 weeks were stained by Alizarin Red to observe matrix mineralization (A-F). Consistently, BMSC constructs were more mineralized than SMSC constructs. For each cell type, large pore constructs show more dark red areas indicating more highly mineralized matrix, compared to small pore constructs. Quantification was performed by threshold image analysis (G, H; n = 10 images/cell type/pore size from at least four independent samples in each group) to compare mineralization. Heavy versus light staining were separated to elucidate the contribution of mineral nodule formation from disperse mineralized matrix, seen most intensely in the case of large pore constructs for both cell types. Small pore constructs significantly restrict mineralization as a functional terminal phenotype of BMSC and SMSC. Scale = 200 um. Values represent the mean ± SD. In the case of both cell types, large pore constructs are significantly (***) more mineralized than medium or small pore constructs.

Scaffold Pore Size Regulates Vascularization

Previously our group reported that a sufficiently small pore diameter (< 125 um) is necessary to prevent robust vascularization in 3D constructs after in vitro culture allowing for cell migration and engineered tissue formation. This is an important design criteria to modulate osteogenic and chondrogenic fate of a cartilage template by preventing or facilitating ossification [19]. In the present study, constructs were implanted subcutaneously only 24 hours after cell seeding. After 4 weeks of cell-scaffold maturation in vivo, 250–425 um pore scaffolds were robustly vascularized throughout the depth of the construct as assessed by CD31+ immunohistochemistry (Fig 5A–D). This phenomenon is consistent after 8 weeks. Scaffolds with 125–250 um pore diameters are well-vascularized by 4 weeks, with total CD31+ area comparable to the 250–425 um diameter constructs. However, 125–250 um pore diameters reach a limit as to their vascularization (Fig 5C”) whereas pore sizes greater than 250 um continue to vascularize up to 8 weeks (Fig 5E). At both time points, constructs with pores less than 125 um in diameter are significantly less vascularized, regardless of cell type. Blood vessels infiltrating into constructs with pores less than 250 um in diameter penetrate the constructs to significantly lower depths than large pore constructs at both time points (Fig 5F). The diameter of blood vessels in these constructs is also significantly lower than in large pore constructs (Fig 5G). Constructs with pores less than 125 um in diameter show the lowest amount of vascularization by CD31+ image area, blood vessel penetration depth, and blood vessel diameter. Notably, these blood vessels are largely Type-H vessels, as demonstrated by immunohistochemistry for endomucin in subcutaneous sections (Fig S3). Therefore, we conclude that even after only 24 hours of culture in vitro (sufficient to allow cell attachment after seeding but before confluency is reached) construct pore size significantly modulates vascularization and vasculature maturity.

Figure 5. CD31 histology + quantification.

CD31+ immunohistochemistry of subcutaneously implanted cell-scaffold constructs at four and eight weeks in vivo (A-D). CD31 expression is quantified to better contextualize the effect of pore size on scaffold construct vascularization patterns in vivo. CD31+ area by section (E), blood vessel penetration depth from the perimeter of the construct (F), and blood vessel diameter (G) are plotted at 4 and 8 weeks, by cell type and construct pore size. In the case of both cell types, small pore scaffolds modulate less total vascularization (CD31+ area), and restrict vasculature maturity (penetration depth and diameter), demonstrating pore size as a critical design motif to modulate vascularization. Image analysis was performed on a minimum of 10 images per cell type per pore size. Scale = 200 um; n = 10 images/cell type/pore size from at least four independent samples in each group. Values represent the mean ± SD. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05.

Scaffold Pore Size Regulates Extracellular Matrix Maturation

At both 4 and 8 weeks, there is uniform distribution of total ECM throughout the constructs (Fig 6A–D”). Quantitatively, there is a higher abundance of green (immature matrix) at 4 weeks compared to the same experimental group explanted at 8 weeks across all cell and pore size combinations (Fig 6E, G–J). A general trend is observed that sufficiently small pore scaffolds (<125 um pore diameter) contain more green and yellow (immature matrix) contribution relative to orange and red (mature matrix), for all groups. Increasing pore diameter is directly related to an increasingly mature matrix. Interestingly, SMSC generally have more immature collagenous matrix across all pore sizes compared to BMSC at the same time point. Additionally, we quantified the total amount of picrosirius red+ (PSR+) area representative of total amount of ECM deposited in the scaffold constructs (Fig 6F).

Figure 6. PSR Histology from SubQ.

Subcutaneous constructs were stained with picrosirius red (PSR) and imaged with a polarized light microscope at 4 and 8 weeks (A-D). Polarized light microscope images were processed in Fiji to determine the relative composition of PSR+ area by color (E, shown in more detail in Fig S6) and total PSR+ area per image (F). Red and orange color indicates mature collagen, and green and yellow color indicates immature matrix components. In the case of both BMSC and SMSC, large pore constructs resulted in secretion of a more mature extracellular matrix as determined by the greater relative contribution of red and orange compared to green and yellow areas after PSR staining, while small pore constructs facilitated a more immature matrix composition. Image analysis was performed on a minimum of 10 images per cell type per pore size. Scale = 200 um; n = 20 images/cell type/pore size from at least four independent samples in each group. Picrosirius red quantification by cell type and time point for each color, shown in violin plots (G, H, I, J). Red and orange color indicates mature collagen, and green and yellow color indicates immature matrix components. Large pore constructs for both BMSC and SMSC, at both time points, facilitate a more robust matrix composition as determined by the greater relative contribution of red and orange compared to green and yellow. On the other hand, small pore constructs for both cell types and at both time points facilitate a more immature matrix with the greatest contributions coming from green and yellow colors. *p<0.05.

Scaffold Pore Size Differentially Promotes Cellular Differentiation

We performed q-PCR analysis for five broad classes of genes: stem cell markers (CD44, Nestin, Axin2), Skeletal Stem Cell Markers (Gremlin, LepR, CD146), Suture Stem Cell Markers (Gli1), Extracellular Matrix (Col1, Col3), Angiogenesis (VEGF) and Osteogenesis (BSP), as well as Chondrogenesis (Sox9, Col2, Aggrecan). Heat maps demonstrating relative gene expression among experimental groups are shown in Fig 7A and 7D with individual bar graphs shown in Fig S4 and S5.

Figure 7. Subcutaneous gene expression.

Subcutaneously implanted cell-scaffold constructs were harvested and subjected to gene expression analysis, from which heat maps were generated at 4 weeks (A) and 8 weeks (D) time points, for genes related to stemness, osteogenesis and chondrogenesis. Consistent across both cell types, large pore constructs tend to facilitate osteogenic differentiation while small pore constructs do not facilitate osteogenesis and rather maintain expression of stem cell and skeletal stem cell markers. These trends are significant at 4 weeks, and become increasingly divergent by 8 weeks. Pearson similarity matrices were calculated for each heatmap to visualize and deduce trends in the data by correlating pore size, cell type, and gene (B, C, E, F), which corroborate with our conclusion, summarized in G. n = 4 samples per cell type/pore size. Values represent the mean ± SD. Additionally we subjected these constructs to immunohistochemistry of CD146+ in histological sections, a skeletal stem cell marker shared by both BMSC and SMSC to confirm the pore size-mediated expression of CD146.

After 4 weeks in vivo, both BMSC and SMSC cultured in different pore size constructs showed markedly different patterns of gene expression. Constructs with pores less than 125 um in diameter facilitated increased expression of classical stem cell and skeletal stem cell markers, Gli1, and Col3, and downregulation of markers associated with osteogenesis. On the other hand, constructs with pores greater than 250 um in diameter exhibited lowest expression of these markers but much higher expression of Col1, VEGF and BSP (Fig 7A, 7D). Interestingly, there were no significant changes in chondrogenic gene expression (Sox9, Col2, Aggrecan) as a function of pore size (Fig S5), which corroborates with our histologic findings (Fig S2). Pearson correlation coefficients were calculated for all interactions by row and column to generate similarity matrices. Similarity matrix analysis indicates that BMSC-Small Pore and SMSC-Small Pore constructs show similar gene expression patterns to each other (Fig 7B, 7E). Correlation by gene (row) shows strong patterns indicated by similarly colored regions composed of multiple points (Fig 7C, 7F). Classical stem cell markers correlate closely to immature ECM and Gli1 SMSC marker, whereas VEGF (angiogenesis), BSP (osteogenic) and Col1 (mature ECM) are inversely related to these genes.

These trends are consistent at both 4 and 8 weeks in vivo. By 8 weeks’ time, both 125–250 um and 250–425 um diameter constructs are more genotypically similar to each other while the 60–125 um pore constructs become more dissimilar, indicating a unique gene expression pattern in the small pore constructs as shown by Pearson correlation coefficients between pore sizes and cell (columns, Fig 7E). A summary of in vivo gene expression data is shown in Fig 7G. Additionally, immunohistochemistry for CD146, a skeletal stem cell marker common to both BMSC and SMSC is shown in Fig 7H, providing histological evidence which corroborates with gene expression data for the maintenance of stemness in small pore constructs. Based on these data we conclude that constructs with spherical pores less than 125 um in diameter provide a unique geometric microenvironment which promotes stemness and maintenance particularly of the SMSC phenotype, distinct from constructs with spherical pores greater than 250 um in diameter, which facilitate increased expression of classic osteogenic and mature tissue markers over time.

Scaffold Pore Size Modulates Gene Expression in the Absence of Subcutaneous Vasculature

We were curious whether the contributions of pore size and vascularization on gene expression and extracellular matrix deposition were mutually exclusive. Scaffolds of the three pore sizes were cultured with BMSC and SMSC for 3 weeks in vitro not allowing for the subcutaneous vasculature that is seen in vivo. A heat map of qRT-PCR gene expression data is shown in Fig 8A (individual bar graphs in S6). In agreement with in vivo findings, significant differences in gene expression were observed as a function of pore size, with small pores favoring maintenance of stemness while large pores facilitating a trajectory towards osteogenic differentiation. Pearson similarity correlation demonstrated that SMSC and BMSC cultured in 60–125 um pore constructs are the most similar to each other, rather than SMSC or BMSC cultured in larger pore constructs, both 125–250 um and 250–425 um (Fig 8B). When individual genes are correlated to each other, we observe strong trends among stem cell, skeletal stem cell, and immature ECM markers, opposite to trends in osteogenic markers, as predicted (Fig 8C). In vitro culture of cell-scaffold constructs showed a smaller effect on mineralization potential after three weeks; however, pore size did still regulate differentiation towards a future osteogenic fate in terms of mineralization (Fig S7). These findings suggest that scaffold pore size modulates the differentiation trajectory and future fate of SMSC and BMSC even in the absence of ingrowing vasculature.

Figure 8. In vitro qPCR.

3D cell-scaffold constructs were cultured in vitro for 3 weeks and their gene expression was analyzed by qRT-PCR, shown as a heat map (A). Persons’ correlations are calculated by column (B) and row (C) to determine trends. In vitro culture resulted in the same modulation in gene expression as a function of pore size, but not cell type, for both BMSC and SMSC. Large pore cell-scaffold constructs were cultured with and without DMOG treatment to mimic hypoxic conditions in vivo in an effort to elucidate the contributions of vasculature and hypoxia separately from physical pore architecture. Differential gene expression by cell type is shown for BMSC (D) and SMSC (E) after 3 weeks of chemical induced hypoxia versus normoxia. A plot of gene expression as the difference between normoxia and hypoxia is shown (F). n = 5 samples per cell type/pore size. Values represent the mean ± SD. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05.

From our results, we hypothesized that in vivo gene expression and tissue fate is in part a function of construct vascularization and resulting hypoxic tension. To further elucidate the contribution of pore size itself, we cultured 250–425 um pore size constructs seeded with BMSC or SMSC either with or without DMOG, a cell permeable prolyl-4-hydroxylase inhibitor which upregulates hypoxia inducing factor (HIF), to mimic hypoxic conditions (Fig S8). For both cell types, hypoxic DMOG treatment causes changes in gene expression patterns for both cell types (Fig 8D–8E), as a result of increased HIF1α and accumulation (Fig S8). DMOG treatment downregulates osteogenic genes significantly, similar to the changes in gene expression observed in small pore scaffolds that lacked vascularization when cultured in vivo. DMOG treatment also caused downregulation of Gli1. Yet, we demonstrated that scaffolds with pore sizes less than 125 um show significantly higher Gli1 expression. Together, our data is indicative of maintenance of stem cell phenotype in vivo and in vitro by pore size. The data also indicate that vascularization regulates some, but not all aspects of pore size-mediated differential gene expression.

Discussion

In the present study, we demonstrated that scaffold macropore size is a primary regulator of mesenchymal stem cell and tissue phenotype, in vivo and in vitro, for both long bone and craniofacial skeletal MSCs. The sugar sphere porogen method lends an ideal platform to study the effect of pore size because it allows for uniform, spherical pores of size chosen by mesh sieves during fabrication (Fig 1) [20], and is optimized to allow for cell migration and nutrient/waste diffusion (Fig S1). Here, we demonstrate that spherical pores less than 125 um in diameter favor maintenance a stem cell phenotype with an immature, avascular matrix, that ultimately prevents premature cell specification towards the functional mineralizing osteoblast fate. On the other hand, spherical pores greater than 250 um in diameter facilitate matrix maturation, robust vascularization, and support a trajectory towards mineralized bone formation and terminal differentiation of skeletal MSCs. A well-defined ideal microenvironment which shows potential to maintain the suture mesenchyme phenotype as a stem cell population is critical towards developing a regenerative medicine approach to treating craniosynostosis through a single surgical procedure and implantable biomaterial construct.

Vascularization and a mature collagen matrix are prerequisite for cranial bone formation; modulation of these two tissue-level parameters is critical in terms of optimizing regenerative outcomes [37, 38]. In the early osteogenic matrix, small bore capillaries invade to provide nutrients and osteoprogenitor cells; enhanced osteogenesis is closely correlated to increased bone mass [39–41]. Tissue level properties in the cranial suture are markedly different from the flanking calvarial bone, which is largely similar to the properties of long bone. Recent literature suggests that blood vessel growth into the suture precedes osteogenesis as the avascular sagittal suture fuses in a mouse model [42]. Additionally, vascularization is used to mark suture synostosis/fusion associated with aging [43]. Zhao et al reports that Gli1+ SMSCs do not show affinity for vasculature [16]. The presence of both CD31 and endomucin in construct vasculature in vivo shown here (Fig S3) indicates type H-vessels, known to mediate the growth of bone vasculature, maintain perivascular osteoprogenitor cells and couple angiogenesis to osteogenesis [44].

Extracellular matrix composition is likewise is an important aspect of stem cell and tissue microenvironments and regulates cellular behavior [45]. Collagen III is highly expressed in the patent suture and its relative expression compared collagen I decreases concurrently with suture fusion in both humans and mice [46]. Type I collagen accounts for 90% of the matrix protein content in bone, with less than 5% collagen types III and V [47]. We show that scaffold pore size applies a selective pressure towards differential extracellular matrix composition where small pore constructs favor an immature matrix and upregulated Col3 gene expression, while large pore constructs favor a more mature matrix and upregulated Col1 gene expression (Fig 6). Our in vitro data suggests that maintenance or loss of stemness is likely tightly regulated by scaffold pore morphology (Fig 8). Smaller pores have steeper curvature compared to large pores which may put additional selective pressure on cells towards maintaining their stemness rather than differentiating [48]. At the cellular level, pore size likely modulates cell-cell and cell-matrix interactions as a means of modulating differentiation trajectory. These results combined with current literature suggest that scaffold-mediated cell and tissue-level properties are responsible for modulating cell fate inside of biomaterial constructs.

Stem cell niches are dynamic microenvironments within organs and tissues that retain a reservoir of undifferentiated stem cells to maintain homeostasis and repair [49]. Loss of stem cell niches results in impaired healing and regenerative capacity in a number of tissues [30]; permanent niche changes are associated with aging and disease progression [49]. Owing to their MSC properties and importance in craniofacial development and homeostasis, recapitulating the cranial suture mesenchymal stem cell niche in vivo and in vitro with a biomaterial scaffold construct which maintains SMSC is meaningful step towards developing a tissue engineering approach to treating craniosynostosis. Hong and Mao described the first attempt of a tissue engineering approach to treating craniosynostosis [50]. Dermal fibroblasts were seeded in a gelatin scaffold between two collagen sponges loaded with recombinant BMP2, which showed radiolucency after four weeks in vivo. Results of this study and ours and serve as a proof of concept for the idea behind a tissue engineered cranial suture construct. Since this work, significant progress in cranial suture biology and an understanding of the process by which craniosynostosis occurs in the disrupted cranial suture mesenchyme lends itself to the design of tissue engineering and regenerative medicine strategies to treat craniosynostosis. Recently Yu et al describe a tissue engineering approach comining a gelatin methacrylate-based (GelMA) hydrogel with Gli1+ MSCs [51]. The authors describe their optimized biomaterial as having a permeability which prevents angiogenesis, to prevent bone formation and suture fusion, and demonstrated that collagen I as a scaffolding matrix is not appropriate for suture regeneration because it attracts osteoprogenitor cells. The authors demonstrate their material’s ability to maintain Gli1+ expression for 10 days in vitro, and to regenerate a cranial suture which normalized intracranial pressure in a mouse model of Saethre-Chotzen syndrome. We believe that pairing exogenous Gli1+ SMSCs with an appropriate biomaterial platform which predictably elicits specific cellular phenotypes over time, as a result of an optimized physical microenvironment, is an important consideration in cranial suture tissue engineering. Given these findings, our study is the first to demonstrate maintenance of the SMSC population using a precisely controlled biomaterial microenvironment, and holds significant promise for future developments.

Mesenchymal stem cells, isolated from long bone, are capable of maintenance, repair, and regeneration of skeletal tissues, and are multipotent, capable of differentiating towards osteogenic, chondrogenic, and adipogenic trajectories [19, 52–54]. Cranial bones have distinct developmental origins from long bone [55]. The cranial suture is a craniofacial growth site, and recently characterized as a potential skeletal stem cell population for craniofacial skeletal tissue [16], responsible for analogous maintenance, repair, and regeneration as BMSCs in long bone. Zhao et al. demonstrated that ablation of the Gli1+ SMSC population leads to skull growth arrest and compromised craniofacial injury repair, and premature loss of this population leads to a craniosynostosis phenotype [16]. In the present study we aimed to identify tissue engineering scaffold design criteria which modulated the initiation and trajectory of cell differentiation, namely pore size. We observed that small pores generally facilitated maintenance of stemness in BMSC and SMSC, while large pores facilitated differentiation. In the case of all pore sizes evaluated, cells infiltrated the scaffolds evenly, proliferated and formed tissue. Stem cell and skeletal stem cell markers are highly expressed in the small pore scaffolds (Fig 7 and 8). We also demonstrated by immunohistochemistry that CD146, a marker of skeletal stem cells, is highly expressed throughout small pore constructs in vivo even after 8 weeks of implantation, but nearly absent in large pore constructs.

Support from studies leveraging mouse models of craniosynostosis demonstrate that cellular changes in differentiation, survival, and renewal within the suture mesenchyme are responsible for the onset of craniosynostosis and its associated developmental defects [9, 56–59]. In particular, our laboratories have developed and studied models of non-syndromic midline craniosynostosis in a mouse model of aberrant BMP signaling [9, 15], and syndromic coronal craniosynostosis (Crouzon Syndrome) in a mouse model of a human FGFR2 mutation [13]. In the current report we have focused our studies on the ability of biomaterial constructs to modulate wild type BMSC and SMSC phenotype. In future work, we are interested in understanding the potential role that scaffold pore size and pore curvature may influence cell signaling pathways such as BMP and FGF signaling taking advantage of genetic models of disease.

The present study represents an exciting method to maintain the vitality and phenotype of the cranial suture mesenchyme in vivo and in vitro, through an intimate understanding of the effects of scaffold pore design on cell and tissue specification. First, we believe the data presented here informs critical design criteria for a future biomaterial-based tissue engineering strategy to regenerate an unfused cranial suture. Secondly, given that macroporous scaffolds seemingly recapitulate critical tissue-level properties of the cranial suture in vitro, these constructs serve as a culture platform to study and model the cranial suture. We demonstrated maintenance of skeletal MSC stemness in scaffolds of uniform small pore sizes, which could potentially serve as an isolated suture tissue compartment. In developing a clinical therapy, it will be important to understand the ability of complex multi-tissue constructs in which distinct pore size zones are able to control stemness and differentiation. While Gli1 is a stem cell marker of the unfused suture mesenchyme, it is also expressed by other tissues including bone cells; the identification of a suture-specific marker with direct relevance to human health would make this model more clinically relevant, and is a topic of active research in the field of cranial suture biology. Additionally, the molecular mechanism by which physical scaffold features regulate stemness and differentiation remains to be further investigated.

Conclusions

Tissue engineering scaffold pore size is a critical regulator of cell stemness and differentiation fate. We demonstrate three-dimensional cell-scaffold constructs which recapitulate aspects of the cranial suture mesenchyme, a calvarial skeletal stem cell population, in vitro and in vivo, using sufficiently optimized scaffold macropores, less than 125 um in diameter. On the other hand, scaffolds of sufficiently large macropores (greater than 250 um in diameter) facilitate differentiation and cell fate specification towards terminal differentiation to become mineralizing osteoblasts in a highly vascularized and mature extracellular matrix. Our findings from in vivo experiments were further confirmed in vitro, leading to the conclusion that scaffold pore size directly modulates cell and tissue fate. The present findings represent the first report of a biomaterial platform which is adequately engineered to tune the fate of cranial suture SMSCs in vivo and in vitro, recapitulating many of the cranial suture’s critical cell and tissue-level properties. We believe this is an exciting proof-of-concept for future studies. Scaffold pore size will be an essential consideration for de novo formation of a tissue engineered cranial suture mesenchyme in future studies to investigate long-term regenerated tissue outcomes and inform a treatment strategy to replace fused cranial sutures in craniosynostosis patients, agnostic of specific mutations which cause disease.

Highlights.

Well-controlled macropore design regulates stem cell differentiation fate in 3-D scaffolds.

Pores > 250 um diameter support osteogenic differentiation, vascularization and mature ECM.

Pores < 125 um diameter support maintenance of stemness, immature ECM and reduced vascularization.

Macroporous scaffolds are able to modulate cell fate in vitro and direct cell trajectory in vivo.