The influence of osteopontin-guided collagen intrafibrillar mineralization on pericyte differentiation and vascularization of engineered bone scaffolds

Abstract

Biomimetically mineralized collagen scaffolds are promising for bone regeneration, but vascularization of these materials remains to be addressed. Here, we engineered mineralized scaffolds using an osteopontin-guided polymer-induced liquid-precursor mineralization method to recapitulate bone’s mineralized nanostructure. SEM images of mineralized samples confirmed the presence of collagen with intrafibrillar mineral, also EDS spectra and FTIR showed high peaks of calcium and phosphate, with a similar mineral/matrix ratio to native bone. Mineralization increased collagen compressive modulus up to 15-fold. To evaluate vasculature formation and pericyte-like differentiation, HUVECs and hMSCs were seeded in a 4:1 ratio in the scaffolds for 7 days. Moreover, we used RT-PCR to investigate gene expression of pericyte markers ACTA2, desmin, CD13, NG2, and PDGFRβ. Confocal images showed that both non-mineralized and mineralized scaffolds enabled endothelial capillary network formation. However, vessels in the non-mineralized samples had longer vessel length, a larger number of junctions, and a higher presence of αSMA⁺ mural cells. RT-PCR analysis confirmed the downregulation of pericytic markers in mineralized samples. In conclusion, although both scaffolds enabled endothelial capillary network formation, mineralized scaffolds presented less pericyte-supported vessels. These observations suggest that specific scaffold characteristics may be required for efficient scaffold vascularization in future bone tissue engineering strategies.

1. INTRODUCTION

Repair and regeneration of bone defects is a prevalent clinical problem worldwide^1,2. Approximately one million procedures for bone repair are performed in the U.S. alone every year³, which represents almost 30% of the total health care expenditure in the country⁴. The development of improved biomaterials for bone regeneration, therefore, continues to be a critical need in both regenerative medicine and dentistry. Bone autografts remain the gold standard materials for critical-sized bone defects. However, autologous bone has significant drawbacks, such as limited availability, high hospitalization costs, multiple surgical procedures and donor-site morbidity. Engineered bone scaffolds, therefore, have been proposed as a viable alternative^1,2. Synthetic materials for bone regeneration, however, lack the inherent microcirculation of autologous bone, which greatly compromises the regenerative process. Therefore, understanding the scaffold characteristics that are relevant for controllable vasculature formation in regenerative scaffolds is imperative for the development of bone tissue constructs that can be utilized in future clinical applications.

It has been proposed that in order to engineer bone scaffold materials that mimic the microenvironmental characteristics of native bone, the organic and inorganic components should be assembled with a nanoscale structure similar to that of the native tissue^5–10. In bone biomineralization, non-collagenous proteins are a critical component to direct calcium and phosphate (CaP) into the interstitial spaces of collagen fibrils, a process which is typically referred to as intrafibrillar mineralization¹¹. Such intrafibrillar mineralization process has long been recognized as a fundamental feature of the interpenetrating collagen/hydroxyapatite nanostructure of native bone, and may provide a critical contribution to the advantageous mechanical properties of mineralized tissues^11–13. A polymer-induced liquid-precursor (PILP) mineralization process was previously developed to elucidate and replicate the process of intrafibrillar collagen biomineralization in-vitro^6,8,9,14. In this process, negatively charged acidic macromolecules, such as poly-aspartic acid¹⁵, fetuin^16,17, milk osteopontin (OPN)⁹, or silica-based materials^18,19, are added to calcium and phosphate solutions in order to sequester Ca and P ions that phase separate into colloids of an amorphous mineral precursor, which has been suggested to have fluidic character^5,6. This amorphous precursor then infiltrates into the interstices of the collagen fibrils²⁰, leading to a high degree of intrafibrillar mineralization in a process that closely mimics the mineralization of human bone²¹. Scaffold materials engineered using such a process have been shown to mimic several of the nanostructural properties of native bone^8,14,22,23, to significantly enhance the osteogenic differentiation of stem cells^8,24–26, and to improve osteoclast activity during scaffold remodeling⁹, thus representing a highly biomimetic scaffold material with great translational potential.

Despite the desirable osteoinductive properties of intrafibrillar mineralized collagen scaffolds, the vascularization capacity of these biomimetic materials remains to be tested. In fact, although vascularization of various CaP materials has been previously reported^27,28, the specific effects of matrix mineralization on the formation of endothelial capillary networks in-vitro has remained poorly understood. It has been suggested that endothelial cells (ECs) require the presence of stem cells to form capillary networks in CaP scaffolds in vitro^29–32. Collectively, these studies have suggested that ECs cultured alone attach, spread and proliferate as monolayers onto CaP scaffolds, with virtually no capillary-like structure formation^29,32. It is known that during vasculogenesis, human mesenchymal stem cells (hMSCs) migrate alongside endothelial cells, and differentiate into mural cells, or pericytes, which regulate the paracrine and angiocrine signals responsible for the stabilization of newly-formed vessels^33–36. Several strategies to optimize the angiogenic potential of CaP scaffolds have been suggested, including variations in the hMSCs:ECs ratio^32,37,38, loading the scaffold with vascular endothelial growth factors (VEGF)^39,40 and testing different CaP composites such as glass⁴¹, silicate^41,38 and silk fibroin^40,42. Although the osteogenic potential of these scaffolds looks promising, little is known about how such materials might influence the differentiation of hMSCs into pericyte-like cells.

Therefore, to address this knowledge gap, here we engineered and characterized intrafibrillar mineralized collagen scaffolds using a PILP biomimetic mineralization method, and co-cultured hMSCs and human umbilical vein endothelial cells (HUVECs) in both mineralized and non-mineralized collagen samples. We then characterized vasculature formation and expression of key pericyte-related markers in these two conditions. Our results suggest that matrix mineralization significantly reduces the differentiation potential of hMSCs into pericyte-like cells in-vitro, which we argue may provide important evidence of the specific microenvironmental conditions that may be required for efficient vascularization of bone-like scaffolds.

2. MATERIALS AND METHODS

2.1. Scaffold Preparation

Mineralized collagen scaffolds were prepared using commercially available bovine type-I collagen foams (2 × 2 × 2 mm) (RCF Resorbable Collagen Foam, cat # DWD3MM, Ace Surgical Supply, Brockton, Ma, U.S.). Scaffold mineralization was obtained by positioning the collagen foams into histological cartridges and immersing the samples in a solution composed of 0.010 M Tris HCl (0.661 g), 0.008 M Tris base (0.097 g) (pH 7.6), 100 μg/ml of osteopontin (OPN) obtained from bovine milk (Lacprodan® OPN-10; Arla Foods Ingredients Group P/S), and 9 mM CaCl₂^.H₂O (Sigma, St Louis, MO). This solution was sonicated for 10 minutes and subsequently diluted with an equal volume of 4.2 mM K₂HPO₄ (Sigma, St Louis, MO), as described previously⁹. Samples were mineralized for 4 d under slow agitation at 37°C. As a positive control, freshly-cut bone slices from the diaphysis of the femur of Rhesus monkey were obtained post-mortem using a wet saw. The bone specimens were rinsed with DI water, immersed in 1% penicillin/streptomycin solution for 24 hours, then abundantly rinsed in DI water, lyophilized, grounded and stored at 4 °C until use.

2.2. Scanning Electron Microscopy

Collagen scaffold structure and morphology were analyzed via scanning electron microscopy. To that end, scaffolds (n = 3) were prepared as described above, cross-sectioned, fixed using a solution containing 2.5% glutaraldehyde in cacodylate buffer for two hours, and lyophilized overnight. Samples were subsequently sputter-coated with gold/palladium and imaged using a FEI Helios Nanolab 660 DualBeam Scanning Electron Microscope at 20.0 kV or a FEI Quanta 200 SEM at 20.0 kV with an attached energy dispersive spectrometer (EDS).

2.3. Fourier transform-infrared spectroscopy

IR spectra from native bone, mineralized scaffolds, and non-mineralized collagen scaffolds were collected using a Nicolet 6700 FT-IR spectrophotometer (Thermo Scientific, Waltham, MA, USA). At least three samples of each group were dehydrated with a gradient series of ethanol and lyophilized overnight. They were then crushed with 100 mg of FTIR-grade anhydrous potassium bromide (KBr) and pressed to prepare a 7-mm pellet. Spectra were collected in wavelengths ranging from 2000 and 400 cm⁻¹ at 2 cm⁻¹ resolution using 32 scans. The amide I band (C=O stretch at 1650 cm⁻¹), used to normalize the spectra, is related to the polypeptide chain conformation of collagen I. Mineral to matrix ratio was calculated from the integrated areas of the phosphate band at 900–1200 cm⁻¹ relative to the amide I band at 1650 cm⁻¹, following published protocols⁴³. Cristallinity index (CI) was calcutated using the following ratio: CI=(A604+A564)/A590.⁴⁴

2.4. Mechanical properties

The compressive modulus of the control and mineralized scaffolds were obtained by performing unconfined compression tests at room temperature using a universal testing machine (MTS Criterion Model 42, MTS Systems Corporation, MN, USA) equipped with a 100 N load cell. Samples of 2 × 2 × 2 mm were hydrated in phosphate-buffered saline (PBS) and stored overnight at 4°C prior to mechanical testing. Compression was performed with a cross-head speed of 1.0 mm min⁻¹ on samples centered on the test platen. The resultant stress was plotted as a function of strain for representative specimens of each group. The compressive modulus was derived from the slope of the initial linear portion of the curve ranging from 0 to 10% strain (n=6).

2.3. Cell Culture

Human umbilical vein endothelial cells transfected with GFP-Lentiviral particles (HUVECs) were obtained from Angio-Proteomie (cat # cAP-0001GFP, Boston, MA). Cells were cultured in endothelial cell growth medium (EGM, cat # cAP-02, Angio-Proteomie) with 5% fetal bovine serum (FBS) and 1% (v/v) penicillin-streptomycin (PS). HUVECs from passage 4–6 were used. Primary cultures of bone marrow human mesenchymal stem cells (hMSCs) (donated by Dr. Brian Johnstone, OHSU Orthopedics) were cultured in low-glucose Dulbecco’s modified Eagle’s medium (DMEM) (Corning Life Sciences, Fisher Scientific Co, Pittsburgh, PA) with 10% FBS and 1% penicillin/streptomycin. hMSCs were used from passage 2–4. All cells were maintained in a humidified incubator (5% CO₂, 37 °C), and the culture media was changed every two days with one cell passage per week.

2.5. Endothelial capillary network formation and hMSC differentiation

To evaluate both the vessel formation and the differentiation of hMSCs on mineralized vs non-mineralized collagen scaffolds, 20 μl of a co-culture of HUVECs:hMSCs in a 4:1 ratio and cell density of 5 × 10⁷ was seeded directly onto the scaffold using a pipette tip (n=6) so the cells could permeate the scaffold pores. After 1 hour the scaffolds were then transferred to a new well plate and covered with 1:1 EGM:DMEM culture medium and were stored in a humidified incubator (5% CO₂, 37 °C). The cell medium was changed every two days during 7 days.

2.6. Immunofluorescence and capillary network analyses

After 7 days in culture, samples were rinsed with phosphate-buffered saline (PBS), fixed with 4% paraformaldehyde (v/v) for 1h, washed with PBS, permeabilized with 0.1% (w/v) Triton X-100 for 10 min, and blocked with 1.5 % (w/v) bovine serum albumin (BSA) for 1 h. After washing with PBS, samples were incubated with primary mouse monoclonal antibody (anti-smooth muscle actin, Abcam, 1:400), overnight at 4°C. Samples were washed with PBS and incubated with secondary antibody (1:250, goat anti-mouse Alexa Fluor 555, red fluorescence, Invitrogen) for 2 h. This was followed by rinsing in 0.1% PBS, staining of the nuclei using a NucBlue (ThermoFisher Scientific, Waltham, MA USA) for 20 min at 37 °C. Samples were imaged using an inverted fluorescence microscope (FL Auto, Evos). To quantify vascular network formation in the scaffolds, 5 images of each scaffold were obtained using a confocal microscope (Zeiss Elyra PS.1, Zeiss International, Oberkochen, Germany). The depth of imaging was 100–300 μm, split into at least 15 Z-stacks. Three-dimensional (XYZ) Z-stacks were converted to 2D TIFF stacks using Fiji. New capillary network formation was measured using AngioTool (National Cancer Institute, NIH). To that end, images were automatically segmented, skeletonized and analyzed according to a set of pre-determined default parameters to quantify vessel length, vessel area, number of junctions, total number of vessel endpoints.

2.7. Quantitative RT-PCR (qPCR)

To access the influence of the intrafibrillar mineralization on the gene expression of pericyte differentiation markers, we conducted experiments as described above (n=6), and on day 7, samples were processed for real time polymerase chain reaction (RT-PCR). Total RNA was extracted using a Trizon-kit (Direct-zol RNA MicroPrep, cat R2061, Zymo Research, Irvine, CA, USA) as recommended by the manufacturer. The RNA quality, concentration and purity was verified using a Nanodrop 2000 Spectrophotometer (Thermo Fisher Scientific Inc., Wilmington, DE, USA). The obtained total RNA was transcribed into complimentary DNA (cDNA) using SuperScript^® III First-Strand Synthesis System for RT-PCR, Invitrogen, Carlsbad, CA, USA) according to manufacturer’s recommendation. The primers for all genes analyzes and respective annealing temperatures are described in Table 1.

Table 1:

Primers for gene expression analyses of pericyte lineage differentiation markers and annealing temperature.

Target gene	Seq (5’->3’)	Anealling temperature
ACTA2 NM_001141945.2	For: AGCGTGAGATTGTCCGTGACAT	52 °C
	Rev: GCGTTCGTTTCCAATGGTGA	52 °C
CD13 XM_005254892.4	For: GGACAGCGAGTTCGAGGGGGA	58 °C
	Rev: AGTGGCCACCACCTTTCTGACA	58 °C
Desmin NM_001927.3	For: TGAAGGGCACTAACGATTCC	52 °C
	Rev: CTCAGAACCCCTTTGCTCAG	52 °C
NG2 NM_001897.4	For: GCTTTGACCCTGACTATGTTGGC	58 °C
	Rev: TCCAGAGTAGAGCTGCAGCA	58 °C
PDGFβ-R XM_011537659.1	For: TGGTGCTCACCATCATCTCC	52 °C
	Rev: CACCTTCCATCGGATCTCGTAA	52 °C

Transcribed cDNAs were amplified for relative quantification in relation to glyceraldehyde 3-phosphate (GAPDH). SYBR^®Green fluorophore (Power^® SYBR^® Green qPCR MasterMix, Applied Biosystems, Framingham, MA, USA) was used at StepOnePlus™ System (Applied Biosystems) device to evaluate the amount of cDNA product as an exponential phase of the amplification reaction. The 2^−ΔΔCT method was used to analyze the relative changes in gene expression from the quantitative RT-PCR experiment.⁴⁵

2.8. Statistics

Statistical analysis was performed using GraphPad Prism 6. The values represent averages and standard deviations. Student’s t-test and one-way ANOVA followed by Tukey post-hoc test were applied with an alpha of 0.05 for all tests.

3. RESULTS

3.1. Scaffold mineralization and mechanical properties

Figure 1A shows SEM images of the fractured surface of the native bone matrix, with highly mineralized collagen fibrils. Fibrils in the control group (Figure 1B) had loosely packed collagen displaying the typical D-banding periodicity of non-mineralized collagen type I⁴⁶. The mineralized samples, on the other hand, (Figure 1C) presented visibly thicker fibrils where the 67-nm periodicity was not seen. A higher magnification view of the mineralized collagen shows individual fibrils that appear to display crystallites preferentially aligned with the long axis of collagen fibril (Figure 1D). The EDS spectra shown in the insets confirmed the presence of calcium and phosphate in the mineralized samples (Figure 1C), consistent with the elemental composition of native bone (Figure 1A), and absent in non-mineralized collagen (Figure 1B).

Figure 1 – Scanning electron micrographs.

Comparison of the microstructure of the collagen from native bone (A), non-mineralized collagen scaffolds (B) and intrafibrillar mineralized collagen scaffolds (C), with insets showing EDS spectra of each. The collagen fibrils are assembled in a randomly aligned fibrillar structure in all groups. The non-mineralized collagen fibrils show the typical banding pattern (B). Mineralized samples (C,D) present thicker fibrils without the banding pattern. The EDS spectrum reveals the calcium and phosphate presence in both native bone (inset A) and mineralized collagen (inset C), but not in the control (inset B). Scale bar (A,B,C) - 1 μm, (D) 300 nm.

FTIR spectra of non-mineralized collagen, mineralized collagen, and native bone samples showed that the typical peak corresponding to alpha-helical structure of collagen at the amide I region (1650 cm⁻¹) is visible in the all groups (Figure 2A). The absorption bands between 900 and 1200 cm⁻¹, corresponding to the PO₄^3- vibration were slightly higher in the bone than in the mineralized samples. The relative ratio of mineral to organic matrix was calculated based on the peak areas of the PO₄^3- vibration to amide I, indicating that the mineralized scaffolds have a mineral:matrix ratio comparable to that of the native bone (Figure 2B). Native bone and the mineralized scaffolds presented well-resolved peaks at 604 cm⁻¹ and 565 cm⁻¹ which could be used to measure the crystallinity index (CI). We used the method of Shemesh et al.,⁴⁴ where narrower peaks exhibit more splitting and therefore yield a higher CI (Figure 2C). The crystallinity index of native bone was higher than that of mineralized scaffolds (2.71 +− 0.21 and 2.44 +−0.05, respectively (Student`s t test). The mineralized collagen had a more prevalent shoulder at 1110 cm⁻¹ for the v₃PO₄^3- peak, which also suggests it had somewhat lower crystallinity.

Figure 2 –

FTIR spectra (500–2000 cm⁻¹) of the non-mineralized collagen scaffolds, intrafibrillar mineralized collagen scaffolds and native bone (A). The infrared vibrations of phosphate (ν₁PO₄³), and collagen (amide I) are labeled (A). The mineral:matrix ratio of bone and mineralized was calculated based on the integrated areas of the phosphate (900–1200 cm⁻¹) and amide I (1650 cm⁻¹) peaks and presented similar values (B). The crystallinity index (CI) of native bone versus mineralized scaffolds, as determined Shemesh et al. method⁴⁴, with CI=(A₆₀₄+A₅₆₄)/A₅₉₀, where Ax is the absorbance (peak height) at wave number × (Student’s t test).

Figure 3A shows a representative stress versus strain curve obtained in compression for collagen scaffolds with and without mineralization. The mineralization process resulted in a significant increase in the deformation resistance of the collagen samples, where the compressive modulus (stiffness) exhibited a 15-fold increase after mineralization (p = 0.001, Student’s t-test) (Figure 3B).

Figure 3:

Mechanical properties of the intrafibrillar mineralized collagen scaffolds as compared with non-mineralized collagen controls. Typical stress strain curve from non-mineralized and mineralized scaffolds (A). Compressive modulus of the groups (B). The mineralization increased the compressive modulus of the scaffolds in a significant way (Student’s t-test, p = 0.001).

3.2. hMSCs differentiation and capillary formation

We investigated the effect of mineralization on the vascularization and hMSC differentiation of scaffolds co-cultured with HUVECs and hMSCs. There were visible differences in the morphology of the capillary-like structures seen in the two groups. While capillaries in the non-mineralized collagen group appeared to be better defined, with visible branches, and with more adjacent αSMA positive cells (Figure 4 A,C,D), HUVECs in the mineralized scaffolds were more spread and had less visible branches. When co-cultured in the mineralized collagen scaffolds, hMSCs exhibited low expression of αSMA, which suggests that mineralization limited the ability of these cells to differentiate into a pericytic-like phenotype (Figure 4 D,E,F). Interestingly, when the HUVECs:hMSCs were co-cultured on top of mineralized collagen hydrogels, the non-mineralized collagen samples exhibited a higher number of αSMA positive cells than the mineralized collagen (Supplementary figure 1 I,J), also suggesting that collagen mineralization inhibited hMSCs differentiation (Supplementary figure 1).

Figure 4 – Confocal images of hMSC differentiation and angiogenesis in a co-culture of HUVECs:hMSCs.

The cells seeded inside the scaffolds were able to develop a vascular network. The capillaries in non-mineralized collagen scaffolds (B,C) were smaller in diameter and better formed than the ones in the mineralized scaffolds (D,E). Also, there were abundance of mural cells expressing αSMA located alongside the vessels in the non-mineralized group (A). In the mineralized group few cells expressed αSMA in a diffuse manner not characterizing a fully pericytic-like differentiation (D). (Scale 400 μm)

These results were consistent with the quantification of capillary formation, as shown in Figure 5, where non-mineralized collagen scaffolds had nearly three times higher total vessel area than the mineralized group (Figure 5A). Similarly, the total vessel length (Figure 5B), number of junctions (Figure 5C), and number of end points (Figure 5D) were all significantly higher in the non-mineralized collagen than in the mineralized group (p < 0.05).

Figure 5 – Vessel analyses.

In the vascularization experiments, both the non-mineralized fibrillar collagen scaffold and mineralized scaffolds displayed vessel formation. The scaffold area with vessels (p = 0.04), total vessel length (p = 0.008), number of connecting vessels (junctions) (p = 0.04) and the number end points (branch points) (p = 0.04) were significantly higher in the control group (Student’s t-test).

3.3. Pericyte differentiation and gene expression

mRNA expression of common pericyte markers, ACTA and NG2³³, were both downregulated in mineralized samples (p < 0.05). Similarly, CD13, a cell surface protein expressed by pericytes, was downregulated in mineralized samples (p < 0.05) (Figure 6 A,B,C). PDGFRβ, a cell membrane receptor expressed by pericytes, had similar gene quantification in non-mineralized and mineralized groups. Desmin, a muscle specific contractile filament that is present in pericytes showed similar gene expression values for non-mineralized and mineralized samples (Figure 6 D,E).

Figure 6 – Gene expression of pericyte markers.

qPCR reaction revealed downregulation of ACTA, NG2 and CD13 genes in the mineralized group (A,B,C) (Student’s test, p < 0.05). PDGFRb and desmin expression in non-mineralized and mineralized collagen scaffolds were similar (D,E) (p > 0.05).

4. DISCUSSION

Engineering of vascularized bone scaffolds with microenvironmental features that approximate the structure and properties of native bone are highly desired, and have long remained a challenge in the field of bone regeneration. In this work we engineered scaffolds that are characterized by intrafibrillar mineralization, which is a nanostructural hallmark of the native bone microenvironment. These scaffolds were then co-cultured with endothelial and mesenchymal stem cells, and collectively, our data suggests that mineralization significantly reduces the differentiation potential of hMSCs into pericytes.

The first goal of this study was to engineer and characterize intrafibrillar mineralized collagen scaffolds with nanostructural properties and mineralization characteristics closely resembling the native bone microenvironment. We used the intrafibrillar mineralization process with OPN as a process directing agent because preliminary tests comparing OPN with other commonly used polymer-directing agents, such as poly-aspartic acid, showed that the resulting mineralization was more homogeneous, rapid and reproducible in the presence of OPN, also in accordance to previous published studies^8,9,12,14. It is known that post-translational modifications of milk-OPN created by thrombin-cleavage generates proteolytic fragments with distinct effects on hydroxyapatite formation and growth^47,48. Such phosphorylated bovine milk-OPN has been formerly demonstrated to promote mineralization, which is in agreement with our findings⁹. The intrafibrillar mineralization observed in our scaffolds has been suggested to be an advantageous characteristic of a bone substitute^6,24,49–51. It has been reported, for instance, that when stem cells interact with a microenvironment that more closely emulates the nanostructure and local properties of native bone, cells proliferate, have increased osteogenic differentiation potential, and are more capable of remodeling the scaffold in vivo^24–26. The scanning electron microscopy analysis of collagen fibrils in our scaffolds showed the disappearance of D-banding periodicity of individual fibrils, an apparent increase in fiber thickness, and the appearance of streaks on the longitudinal axis of the fibrils, which are all consistent hallmarks of intrafibrillarly mineralized collagen. It has been extensively reported that non-collagenous proteins and synthetic protein analogues can guide intrafibrillar mineralization via formation of amorphous calcium phosphate (ACP) that enters into the gap zones of collagen fibrils, wherein the bed of ACP begins to crystallize, eventually giving rise to elongated electron-dense crystals within the fibril. Such a process has been shown to result in the expansion of individual fibrils perpendicular to their long axis, in a way that the fibril becomes deformed by the mineral^9,11,50, masking the banding pattern, which is consistent with the increased diameter of fibrils we see in our mineralized scaffolds (Figure 1 C,D). The mineral characterization of the scaffolds using FTIR showed the presence of calcium phosphate peaks mainly at 900–1200 cm⁻¹, which is consistent with the formation of hydroxyapatite crystals (Figure 2A)^43,52,53. The EDS spectra also confirmed the presence of Ca and P peaks seen in the freshly extracted bone (Figure 1A, insets). The calculated mineral to matrix ratio of the native bone and the mineralized scaffolds were also comparable (Figure 2B), supporting the biomimetic characteristic of these scaffolds.

It has been shown that pre-vascularized scaffolds can improve the survival of bone substitutes^54,55. However, the formation of functional blood vessels in engineered scaffolds requires the interaction of endothelial cells with differentiated stem cells that express pericyte-like behavior. To test our hypothesis that osteogenic scaffolds may significantly influence the potential of hMSCs to differentiate into pericyte-like cells, we co-cultured HUVECs and hMSCs in our biomimetically mineralized scaffolds and compared against cells cultured in non-mineralized samples. There have been a number of studies that achieved intrafibrillar mineralization of bone scaffolds, but these have mostly focused on osteogenic differentiation of loaded cells^24,49,51. Antebi et al.,⁸ fabricated mineralized scaffolds using dynamic intrafibrillar mineralization directed by polyaspartic acid and then cultured hMSCs into the scaffolds. As a result, mineralized collagen composites were similar to trabecular bone in both ultrastructure and composition, biocompatible for hMSCs, and able to being remodeled by these cells. Ye et al.,²⁴ for instance, used sodium tripolyphosphate and a low molecular weight polyacrylic acid to mineralize collagen, obtaining intrafibrillar and extrafibrillar hydroxyapatite mineral via a bottom-up assembly process. Human umbilical cord mesenchymal stem cells (hUCMSC) were then seeded into the mineralized scaffolds showing good cytocompatibility, adhesion, proliferation, and osteogenic differentiation²⁴. Another attempt to achieve biomimetic mineralization of collagen scaffolds for stem cell differentiation was performed by Harding et al.,⁴⁹, and it was found that a higher amount of extrafibrillar mineralization contributed to a reduction in cell proliferation. On the other hand, the apatite mineral induced high levels of RUNX2 expression⁴⁹. A similar method using carboxymethyl chitosan as the negatively charged polymer was used by Wang et al. to mineralize collagen scaffolds, and it was observed that intra- and extra-fibrillar mineralization stimulated MC3T3-E1 cells proliferation and differentiation⁵¹.

Regarding the formation of capillary-like networks and differentiation of the hMSCs into a pericyte-like lineage, first, we performed preliminary tests to optimize the ratios of hMSCs and HUVECs, and our results showed no difference in α-SMA expression among ratios of 1:1, 4:1, 1:4⁵⁶. Since our objective was to form interconnected endothelial cell networks that were supported by pericyte-like cells, we chose HUVEC:hMSC at a ratio of 4:1, as to allow robust endothelial cell morphogenesis based on recent reports^28,57. Second, we used 1:1 medium because we wanted to give equal chance for cells to retain their vasculogenic and stem cell phenotypes. Having more vasculogenic medium could favor more pericyte differentiation, while having more stem cell medium could risk HUVECs not expressing their phenotype. Therefore, we chose the 1:1 composition for this study. Importantly, other studies asking similar questions used vasculogenic medium only^27,57, and indeed a greater level of pericyte differentiation was found, which we believe to be influenced by the medium composition, and not so much the matrix microenvironment.

It is noteworthy that our experiments were performed in mineralized scaffolds that are comprised of large pore sizes, which allowed cells to infiltrate and spread upon seeding. Therefore, different from typical cell-laden hydrogels where cells spread in 3D, cells in our scaffolds were more likely to spread as a monolayer onto the mineralized scaffold walls.⁵⁸ After a few days, the cell-cell interactions would give rise to the remodeling process required for tubulogenesis and vasculature formation. Figure 4 suggests that the non-mineralized scaffolds had more abundant capillary-like structures, characterized by the elongated and interconnected cells forming ring-like structures generally surrounded by αSMA positive cells. In the mineralized scaffolds, instead, endothelial cells appeared interconnected but were visibly more spread, and supported by faint αSMA-expressing cells.

Other possible reasons for the spreading of endothelial cells inside mineralized collagen is that extracellular calcium is known to promote adhesion, chemotaxis and tubulogenesis of endothelial cells^59,60. Interestingly, the HUVECs in mineralized samples were spread, with fewer tubular formation than non-mineralized collagen. We hypothesize that while the non-mineralized collagen is more easily remodeled during vascular morphogenesis, mineralized samples would require much more rigorous extracellular matrix processing to remodel it^61,62, which may have prompted endothelial cells and hMSCs to spread, as opposed to forming cords and undergoing tubulogenesis. In a prior study using OPN PILP-mineralized collagen scaffolds, a marked increase in osteoclast activity and formation of sealing zones was thought to result from the RGD ligand present in OPN⁹. Perhaps in our studies the RGD ligand could have enhanced the spreading of the endothelial cells; thus, one has to be cautious in interpreting our results as only being caused by the presence of mineral.

We argue that other possible reason for such different endothelial cell morphology is that the calcium phosphate rich environment may interfere with the hMSCs differentiation into pericytes and the final remodeling of microvessels. Previous studies have shed light on the molecular pathways involved in cellular uptake of extracellular Ca²⁺ and PO₄^{3- 63,64}. Extracellular Ca²⁺ signaling has been implicated in the upregulation of osteogenic genes, specifically bone morphogenetic protein 2 (BMP-2), osteocalcin, bone sialoprotein and osteopontin through MAPK signaling⁶³. Similar osteogenic stimulation has been reported in hMSCs in the presence of extracellular PO₄^{3- 64}. These studies suggest that the highly mineralized microenvironment may limit pericyte differentiation due to the presence of mixed osteogenic and vasculogenic signals, which may inhibit stem cell fate commitment to pericytic phenotypes.

In addition to these biochemical signals, stem cell behavior and fate are influenced by the biophysical properties of the cell microenvironment, with matrix stiffness playing a significant role in specification of stem cell differentiation^65,66. These biophysical cues are transmitted to the cells through integrin mediated regulation of cytoskeletal tension with stiffer matrices being more prone to osteogenic differentiation induction^67–69. Our data indicates that mineralized scaffolds are significantly stiffer (Figure 3) than the collagen matrix owing to the reinforcement of the collagen fibrils with hydroxyapatite mineral. Previous reports in our lab however, support the conjecture that pericytic differentiation is favored in relatively softer matrices (1.5 kPa)⁷⁰. This could offer yet another explanation for the diminished number of αSMA⁺ cells in mineralized scaffolds relative to the non-mineralized ones. This seems reasonable given that in the case of bone tissue, vasculature is laid down during bone formation by the multicellular unit, and not added to an already stiff mineralized bone matrix.

It should be noted that one limitation of our methods was the difficulty in imaging the mineralized collagen using the laser-confocal microscope, because the mineralized lattice scatters the light beam and creates more background, making it harder to identify αSMA expression in the individual cells. Nevertheless, the differences were significant enough to have confidence in the results

An additional aspect that deserves further discussion is that our quantification of capillary formation (Figure 5) suggests that endothelial networks were significantly more enhanced in non-mineralized scaffolds. It has been well documented that endothelial cells secrete paracrine and angiocrine factors that can enhance pericyte-differentiation when co-cultured with cells from mesenchymal origin^35,36. Therefore, the increased formation of vascular capillaries in the non-mineralized collagen group may have further enhanced the expression of αSMA, whereas mineralized scaffolds downregulated gene expression of pericyte markers (Figure 6).

While we investigated possible mechanisms that contribute to the inhibition of pericyte differentiation and vasculature formation in this manuscript, more detailed mechanistic studies would be required to identify the specific signaling pathways that determine pericyte lineage specification, and how these are affected in the mineralized scaffolds. Ultimately, possible future strategies to improve vascularization of engineered bone scaffolds for tissue regeneration may be controlled spatial or temporal presentation of pericyte differentiation factors to cells within scaffolds, such as PDGF-BB.

5. CONCLUSIONS

In conclusion, biomimetically mineralized collagen scaffolds enabled endothelial capillary network formation, but decreased the differentiation of hMSCs into a pericytic lineage. Feasibility for vascularized constructs has been demonstrated, but further strategies to enhance the formation of pericyte-supported vascular capillaries in mineralized scaffolds in-vitro may be warranted. This work provides a new platform for examining the effects of matrix mineralization on the formation of pericyte-supported vascular capillaries in engineered tissue constructs. These observations may be relevant for the engineering of vascularized bone scaffolds for future clinical applications.