Induced Pluripotent Stem Cell-Derived Endothelial Networks Accelerate Vascularization But Not Bone Regeneration

Abstract

Vascularization is critical for engineering mineralized tissues. It has been previously shown that biomaterials containing preformed endothelial networks anastomose to host vasculature following implantation. However, the networks alone may not increase regeneration. In addition, a clinically applicable source of cells for vascularization is needed. In this study, vascular networks were generated from endothelial cells (ECs) derived from human induced pluripotent stem cells (iPSCs). Network formation by iPSC-ECs within fibrin gels was investigated in a mesenchymal stem cells (MSCs) coculture spheroid model. Statistical design of experiments technique was evaluated for its predicting capability during the optimization of experimental parameters. The prevascularized units were combined with hydroxyapatite nanoparticles to develop a vascularized composite hydrogel that was implanted in a rodent critical-sized cranial defect model. Immunohistological staining for human-specific CD31 at week 1 indicated the presence and maintenance of the implanted vessels. At 8 weeks, the prevascularized systems resulted in higher vessel density over MSC-only scaffolds. The implanted vessels appeared to establish flow with host vasculature. While there was a slight increase in bone volume in the prevascularized bone construct compared to MSC-only bone constructs, there was not a profound increase in bone regeneration. These results show that scaffolds with network structures can be generated from ECs derived from iPSC and that the networks survive and inosculate with the host postimplantation in a bone model.

Impact statement

Vascularization is critical for engineering bone. Prevascularized scaffolds have been shown to improve postimplantation vascularization. Herein, vascularized networks were generated from induced pluripotent cells derived from endothelial cells. These vascularized units were combined with a fibrin/hydroxyapatite scaffold to develop a prevascularized construct for bone regeneration. Implantation of these scaffolds in a small animal cranial defect model resulted in network inosculation and increased vascularization, but exhibited only a limited effect on bone formation. This study provides insight into the challenges of generating vascularized bone.

Introduction

Vascularization is pivotal to bone growth, development, healing, and remodeling and therefore is a prerequisite for bone tissue engineering. Endochondral ossification that transforms cartilage to bone cells requires close interaction between cartilage and vasculature.¹ In addition, vascularization is equally important for intramembrane ossification, however, the relationship between vascularization and intramembrane ossification is yet to be fully established.² Moreover, impaired vascularization or altered interaction between vasculature and bone cells can disrupt bone growth and may result in various clinical manifestations.³ Establishing functional vasculature is a critical bottleneck to the clinical translation of tissue engineering strategies, including bone tissue engineering.^4–7 In fact, a lack of vascularization is a major barrier in in vivo translation of bone tissue engineering.⁸ Current strategies for vascularization include growth factor delivery, cell transplantation, prevascularization of scaffolds, or some combination of these approaches.^9–14 Prevascularized bone construct may be able to promote integration and survival of bone tissue through connection with host vasculature. Prevascularized materials formed from the coculture of endothelial cells (ECs) and support cells, typically mesenchymal stem cells (MSCs), can inosculate with host vessels and enhance perfusion postimplantation.^15–17 However, these studies often apply “model” ECs, including cells isolated from human umbilical or saphenous veins serving as an important “proof-of-concept,” but not clinically viable, solution.¹⁸

Induced pluripotent stem cells (iPSCs) are a potential source of autologous cells for applications in tissue engineering and regenerative medicine.¹⁹ iPSCs are pluripotent stem cells that can be derived from fibroblasts or peripheral blood from adult human donors. iPSCs exhibit self-renewal and methods have been developed for differentiation into ECs with high efficiency and reproducibility.²⁰ Adult endothelial progenitor cells isolated from peripheral blood are a potential option but exhibit age and disease dependent alterations in function which may limit application.^21,22 However, iPSCs-ECs isolated from patients both with and without the disease have been shown to develop functional vascular networks following in vivo implantation in mice.²³ The proangiogenic potential of iPSCs has been demonstrated in a variety of models, including in vitro vascularization models,^24,25 within synthetic hydrogels,^26,27 for treatment of peripheral artery disease^4,28 and wound healing.²⁰ However, relying on vascularization that occurs from cell assembly or stimulation post implantation maynot be sufficient for engineering tissues of clinically relevant sizes. Implants dependent on postimplant vascularization often fail due to a lack of integration and inadequate vascularization.⁷ Generating microvascular network structures before implantation may allow for rapid vascularization to promote survival of bone cells and implant tissue integration.

A significant challenge in the design of any tissue engineering strategy is the identification of optimal conditions from a typically large parameter space. Exploring the full range of variables (e.g., cell number, cell ratios, media conditions, scaffold properties) in traditional experimental methods would be both incredibly time-consuming and costly. In most cases, experimentalists select a smaller space based on practical considerations and a priori knowledge. However, statistical design of experiments (DOE) techniques can be used to systematically explore complex multivariable systems with greater precision while minimizing the number of experiments necessary to model the system.^29,30 We evaluated DOE techniques as a tool to model optimal parameters for tissue engineering applications.

Biomaterial scaffolds containing extensive networks formed from iPSC-ECs may increase the volume of bone that can be engineered. Hydrogels with networks formed from human umbilical vein endothelial cells (HUVECs) and MSCs have been shown to establish flow following implantation in a rodent cranial defect model, but it was not clear that the presence of functional vessels enhanced bone formation.¹⁷ Hydroxyapatite (HA) has been extensively used in bone tissue engineering applications for its bioactivity and osteoconductivity.^31,32 HA nanoparticles integrated biomaterial scaffolds resulted in enhanced bone formation in various studies.^33–37

In this study, vascularized bone constructs were formed in composite HA/fibrin scaffolds and examined in a rodent critical-sized defect model.³⁸ In addition, DOE techniques were evaluated for optimization of the vascular network formation from iPSCs-ECs/MSCs cocultures. These results demonstrate, for the first time to our knowledge, the survival and functionality of human iPSC-ECs vascular networks following implantation. Our findings suggest that vascular networks formed with iPSC-ECs may be a clinically feasible option for enhancing vascularization and investigating their role in intramembrane ossification.

Materials and Methods

Cell culture

Human iPSC-ECs were purchased from Cellular Dynamics, cultured according to the manufacturer's instructions in iPSC-EC growth medium, and used at passage 3 for all studies. The iPSC-EC medium was prepared as recommended by the cell manufacturer, which consisted of VascuLife VEGF Endothelial Medium Complete Kit (Lifeline Cell Technology, Frederick, MD) with 4 mM l-glutamine and supplemented with Cellular Dynamics Maintenance Medium Supplement (Cellular Dynamics, Madison, WI) in place of fetal bovine serum. Human MSCs (bone marrow-derived) were purchased from Lonza (Walkersville, MD), cultured in Mesenchymal Stem Cell Growth Medium (Lonza), and used at passages 4–5. Upon reaching confluence, cells were washed with phosphate-buffered saline (PBS), detached from the flask using TrypLE (Life Technologies), and cultured to form spheroids as described in section “Fibrin gel formation and spheroid encapsulation” below.

Statistical DOE and three-dimensional culture

A statistical DOE approach was performed using MODDE 12 Software (Umetrics, Sweden) to optimize parameters for maximum vessel-like network formation in three-dimensional (3D) coculture conditions. Two statistical design models were created to directly compare the traditional experimental design with a DOE approach. The traditional experimental design was modeled using three-level full factorial design and fit with partial least squares with experiments defined in Table 1 (45 conditions in triplicate for 135 total samples). A second model was created using a d-optimal quadratic design using 17 total samples recommended by the software, detailed in Table 2. For all statistical design models, three parameters (percent MSCs, fibrinogen concentration, and culture time) were varied and four responses (MSCs outgrowth area, EC outgrowth area, vessel length, and diameter) were evaluated. These parameters were selected based on our previous studies.¹⁷ Parameter constraints were guided by results from preliminary studies (data not shown).

Table 1.

Experimental Conditions for Full Factorial Design of Experiments Model and In Vitro Study

Fibrinogen concentration (mg/mL)	Percent MSC in spheroid (%)	Culture time (days)
1, 3, 5	0, 25, 50, 75, 100	7, 14, 21

MSC, mesenchymal stem cell.

Table 2.

Sample Conditions Defined by Software in Design of Experiments d-Optimal Model

Sample number	Percent MSC (%)	Fibrinogen concentration (mg/mL)	Culture time (days)
1	0	1	7
2	0	1	21
3	0	3	14
4	0	5	7
5	0	5	21
6	25	5	21
7	50	1	14
8	50	3	7
9	50	3	14
10	50	3	14
11	50	3	14
12	75	5	21
13	100	1	7
14	100	1	21
15	100	3	21
16	100	5	7
17	100	5	14

Fibrin gel formation and spheroid encapsulation

Spheroids were formed by incubating varying ratios of iPSC-EC and MSCs (defined as 0–100% MSCs, 5000 total cells per spheroid) overnight in 150 μL iPSC-EC medium with 0.25% wt/vol methylcellulose in a nonadherent, round-bottom 96-well plate.³⁹ Fibrin gel scaffolds were formed by combining 80 μL each of human fibrinogen (2–10 mg/mL; Sigma-Aldrich, St. Louis, MO) and bovine thrombin (2–10 U/mL; Sigma-Aldrich) in a 48-well plate, for the final fibrin concentrations of 1–5 mg/mL. Spheroids were placed on top of the base fibrin layer (one spheroid per scaffold) and encapsulated in a top layer of 80 μL each fibrinogen and thrombin, for a final total volume of 320 μL. Samples were incubated in the iPSC-EC growth medium at 37°C and 5% CO₂ for 7, 14, or 21 days. Following incubation, samples were fixed in 10% phosphate-buffered formalin (Fisher Scientific, Pittsburgh, PA) for 20 min for subsequent analysis.

Immunofluorescent staining

Immunofluorescent staining for the EC marker CD31 and the MSC marker CD90 was performed to assess iPSC-EC/MSC network formation. Following fixation, samples were washed in PBS and blocked with 6% normal goat serum (MP Biomedicals, Solon, OH). Samples were treated with 1:200 mouse anti-human CD31 (clone JC70A; Dako, Carpenteria, CA) and/or 1:100 rabbit anti-human CD90 (Abcam, Cambridge, MA) primary antibody with 2% normal goat serum in 0.5% Triton X-100 in PBS (PBST) at 4°C overnight. Following incubation, samples were washed in 0.5% PBST and incubated with 1:200 tetramethylrhodamine-conjugated goat anti-mouse immunoglobulin G (IgG) and/or 1:200 fluorescein isothiocyanate-conjugated donkey anti-rabbit IgG secondary antibody (Jackson ImmunoResearch, West Grove, PA) for 4 h at room temperature. Samples were imaged with a Zeiss LSM 5 PASCAL confocal microscope. Two-dimensional projections of the 3D z-stack images were created using Pascal software (Carl Zeiss, Germany). Mosaic images were stitched together using the MosaicJ plugin for ImageJ (NIH).^18,19 The area of the CD31-positive cellular network and the length and diameter of individual vessel-like structures were manually outlined and measured for all samples using Axiovision AC 4.5 (Carl Zeiss). These values were input into both DOE models, and surface plots were created to visualize the predictive models.

Fibrin-HA composite scaffold implant preparation

Following the completion of the in vitro study, the optimal conditions determined by the DOE models were used for in vivo studies. Five thousand cells (60% MSCs and 40% iPSC-EC) were suspended in 150 μL iPSC-EC growth medium with 0.25% wt/vol methylcellulose in each well of a nonadherent round-bottom 96-well plate and incubated overnight to form spheroids. MSC-only spheroids were prepared the same way, with 3000 cells per spheroid (100% MSCs). Fibrin gel scaffolds were formed by combining 30 μL each of human fibrinogen (6 mg/mL; Sigma-Aldrich) and bovine thrombin (6 U/mL; Sigma-Aldrich) in a 96-well plate, for a final fibrin concentration of 3 mg/mL. Spheroids were placed on top of the base fibrin layer (two per scaffold) and encapsulated in a top layer of 30 μL each fibrinogen and thrombin, for a final total volume of 120 μL. Samples were incubated in the iPSC-EC growth medium (refreshed every 2–3 days) at 37°C and 5% CO₂ for 21 days.

Composite hydrogels were formed by suspending four cultured fibrin scaffolds in a HA–fibrin mixture (Fig. 1). Two hundred milligrams of nanocrystalline HA (Sigma-Aldrich) was suspended in 1 mL PBS and vortexed for 5 min. Four prevascularized scaffolds were removed from the wells and placed into a cylindrical mold (9 mm diameter, 3 mm height). One hundred microliters of the HA–PBS mixture was combined with 100 μL fibrinogen (12 mg/mL) and 200 μL thrombin (6 U/mL). Final concentrations were 3 mg/mL fibrinogen and 3 U/mL thrombin with 50 mg/mL HA. The mixture was pipetted into the mold and gently mixed to surround the prevascularized fibrin units. Composite scaffolds were incubated in their molds at 37°C and 5% CO₂ in iPSC-EC growth medium overnight before implantation.

FIG. 1.

(A) Schematic depicting production of prevascularized composite hydroxyapatite/fibrin scaffolds. Spheroids are encapsulated in fibrin scaffolds in a 96-well plate and cultured for 3 weeks. Before implantation, four prevascularized fibrin scaffolds are placed in a mold and suspended within hydroxyapatite-loaded fibrin. (B) Implanted prevascularized composite hydroxyapatite/fibrin scaffolds. Color images are available online.

Scanning electron microscopy imaging

Scanning electron microscopy (SEM) was used to evaluate the structure of HA–fibrin composite scaffolds. Fibrin-only and HA–fibrin scaffolds were incubated in 2.5% glutaraldehyde at 4°C for 2 h and washed three times with deionized water. Scaffolds were placed in 5 mL Eppendorf tubes, flash-frozen in liquid nitrogen, and lyophilized overnight. Scaffolds were deposited on carbon conductive tape adhered to SEM aluminum pin stubs and imaged using a Phenom PRO Desktop SEM (Phenom-World, Netherlands) operated at 10 kV.

Cranial defect surgeries

All animal experiments were performed at Edward Hines, Jr. VA Hospital in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) (protocol #H17-002). Cranial defect surgeries were performed on 12 male athymic RNU nude rats (Charles River Laboratories, Wilmington, MA) aged 13–14 weeks according to the established procedure.²⁰ In brief, rats were anesthetized using 3% isoflurane and placed in a small animal stereotaxic instrument (Kopf Instruments, Tujunga, CA) to stabilize the skull. Five hundred microliters lidocaine (1%) was injected subcutaneously as a local analgesic. A 3 cm incision was made longitudinally along the midline of the skull and the periosteum was preserved and pulled back to expose the calvarium. An 8 mm diameter full-thickness defect was created using a trephine bur under saline irrigation. One scaffold was placed in each defect. The periosteum was sutured closed using 5-0 resorbable sutures, and skin was sutured using 4-0 nylon sutures. Buprenorphine-SR (1.0 mg/mL, 1.0 mg/kg) was injected subcutaneously in loose skin anterior to the shoulder as a long-term (72 h) analgesic. Drops of cefazolin were administered along the suture line to prevent infection. Animals were sacrificed at 1 week to assess the survival and anastomosis of the implanted vascular networks, and at 8 weeks to assess vascularization and bone regeneration (n = 4 rats per group per time point). Just before sacrifice, AlexaFluor 647 conjugated isolectin (100 μg; Invitrogen, Carlsbad, CA) was injected via tail vein and circulated for 10 min, followed by euthanasia via CO₂ asphyxiation and perfusion fixation with 10% phosphate-buffered formalin. Harvested samples were formalin-fixed and prepared for microcomputed tomography (μCT) imaging. Following μCT, samples were decalcified overnight in Cal-Ex decalcifying solution (Fisher Scientific). Decalcified samples were cut in half along the sagittal midline; one half was frozen in optimal cutting temperature for frozen sections and one half was paraffin-embedded for histological analysis.

μCT imaging and analysis

μCT analysis was performed as previously described.²¹ In brief, the samples were scanned while hydrated with formalin using μCT SkyScan 1076 (Bruker) at a 100 kV source voltage and a 100 μA source current with a 0.5 mm aluminum filter used and at a spatial resolution of 8.77 μm. The reconstructions were performed using NRecon software (Bruker), and this resulted in grayscale images corresponding to grayscale values from 0 to 255. DataViewer (Bruker) was used to reslice the μCT images along coronal and sagittal axes to reorient the μCT slices to be perpendicular to the cranial-caudal axis of the calvaria, and μCT thresholding was performed to include only ossified tissues using the Otsu algorithm as previously described.²² The threshold for all samples in the study was determined to be 42 on a scale of 255. The region of interest (ROI) was chosen by creating a volume that spanned 8 mm around the defect generated in the calvaria and repaired with allograft to include the entire defect generated. Additional analyses were performed by choosing concentric ROI 7 mm and 9 mm in diameter as well as the remainder of the native calvaria. All ROI comprised the full thickness of the calvaria. The total volume of bone in this 3D volume was computed using CTAn software (Bruker) from the binarized data. The mean bone density of the ossified tissues within these 3D volumes was calculated before binarization as the mean grayscale value of all pixels greater than the chosen threshold. Three-dimensional reconstructions were performed using the Mimics software package (Materialise, Leuven, Belgium) as previously described.²³

Histological analysis

Hematoxylin and eosin (H&E) and Masson's trichrome staining were performed on paraffin-embedded sections (10 μm thickness) prepared after μCT imaging to visualize tissue structure. Stained sections were imaged using an Axiovert 200 inverted microscope (5 × objective, 1.10 μm/pixel), and areas of new mature and immature bone formation within the defect were manually outlined and quantified as a percent of total tissue area within the defect. Four nonconsecutive sections were analyzed per sample using Axiovision software.

Immunohistochemical staining

Immunohistochemical staining for the EC marker CD31 was performed to visualize vascularization within the defect area. Paraffin-embedded sections (10 μm thickness) were deparaffinized, rehydrated, and underwent antigen retrieval by incubating in target retrieval solution (Dako) for 30 min in a steamer. Sections were blocked with 5% normal goat serum and incubated with 1:700 rabbit polyclonal CD31 (Santa Cruz Biotechnologies, Santa Cruz, CA) at 4°C overnight. Secondary antibody staining was performed using a Vectastain Elite ABC kit (Vector Laboratories, Burlingame, CA) according to the manufacturer's protocol, and visualized with 3′-diaminobenzidine. Positive staining was confirmed based on vessel morphology and quantified as the total number of vessels per tissue area.

Human-specific CD31 immunohistochemical staining was performed using the same protocol detailed above, substituting a human-specific CD31 primary antibody (1:50 dilution, clone JC70A; Dako) and incubating overnight in a humidified chamber at room temperature.

Statistical analysis

All data are presented as mean ± standard deviation. Statistical significance between groups was determined with one-way analysis of variance (ANOVA) or two-way ANOVA (bone regeneration histological analysis only) using GraphPad Prism 6.0. Tukey's post hoc test was used to compare differences between groups. In all comparisons, p < 0.05 was considered statistically significant.

Results

Optimization of iPSC-EC/MSC network formation

The optimal conditions for vascular network formation were investigated by varying the ratios of iPSCs-ECs and MSCs (defined as percent MSCs), fibrin concentration, and culture time. To evaluate DOE as a predictive tool for optimization of experimental parameters, the results from these initial experiments were fed into DOE models to narrow down the design space expected to achieve maximum EC network formation for comparison with traditional experimental design. The optimal culture conditions determined by the in vitro studies were then utilized for the in vivo evaluation of prevascularized composite hydrogels as described below.

Growth of vascular networks in vitro

Coculture spheroids of human iPSC-ECs and human MSCs were suspended in fibrin scaffolds and cultured in the iPSC-EC growth medium. Culture conditions were varied with respect to cell ratio (defined as percent MSCs), fibrin concentration, and culture time. MSCs outgrowth was visualized using brightfield microscopy. MSCs growth was observed to increase with time (Appendix Fig. A1) and decreased with an increase in fibrin concentration (Appendix Fig. A2). The highest MSCs outgrowth was observed in the 50% MSCs spheroids encapsulated in 1 mg/mL fibrin (79 ± 2 mm²). The outgrowth was significantly greater when compared to 25% MSCs in 1 mg/mL fibrin (55 ± 10 mm², p = 0.0032), 50% MSCs in 3 mg/mL fibrin (53 ± 8 mm², p = 0.0018), and 50% MSCs in 5 mg/mL fibrin (46 ± 1 mm², p < 0.0001) conditions.

Immunofluorescent staining for the EC marker CD31 was performed to assess iPSC-EC network formation. Cocultured iPSC-ECs and MSCs exhibited interconnected networks of vessel-like structures with 3 or 5 mg/mL fibrin. Spheroids encapsulated in 3 mg/mL fibrin scaffolds demonstrated a larger network area compared to those encapsulated in 5 mg/mL scaffolds (Fig. 2B). In the 1 mg/mL fibrin scaffolds, however, no EC sprouts or vessel-like structures were observed regardless of spheroid cell ratios. The ECs either migrated outward to form flat sheets (Fig. 2E) or formed thin sprouts with no visible lumen or 3D sprouting (Fig. 2F).

FIG. 2.

iPSC-EC network formation is affected by cell ratio and fibrin concentration. Confocal microscopy images of immunofluorescent staining for CD31 (iPSC-EC, red) show formation of vessel-like networks in (A) 25% MSC, (B) 50% MSC, and (C) 75% MSC spheroids encapsulated in 3 mg/mL fibrin and cultured for 3 weeks. (D) No network formation was observed in the absence of MSCs (0% MSC group). (E) Spheroid encapsulated in 1 mg/mL fibrin formed flat sheets expanding through the gel. (F) Spheroid encapsulated in 1 mg/mL fibrin formed thin sprouts with no visible lumen. In all images, scale bar represents 200 μm. EC, endothelial cell; iPSC, induced pluripotent stem cell; MSC, mesenchymal stem cell. Color images are available online.

The extent of iPSC-EC network formation was also dependent on the ratio of iPSC-ECs and MSCs. Vessel-like structures were not observed in spheroids with only iPSC-ECs (0% MSC group). In the absence of MSCs, spheroids separated into individual cells as shown in Figure 2D. At all coculture ratios, spheroids encapsulated in 3 mg/mL scaffolds exhibited increasing network area over time (Appendix Fig. A2B). The 50% MSCs and 75% MSCs spheroids exhibited similar EC network areas after 3 weeks of culture, with the highest in the 3 mg/mL fibrin (50% MSCs: 2.45 ± 0.16 mm²; 75% MSCs: 2.53 ± 0.48 mm²). The length and diameter of individual vessel-like structures were quantified (Appendix Fig. A2). Vessel length and diameter were higher in the 3 mg/mL fibrin conditions than in 5 mg/mL. After 3 weeks of culture, the 75% MSCs group in 3 mg/mL fibrin had the highest mean vessel length of 969.6 ± 79.3 μm and the highest mean vessel diameter of 29.2 ± 3.3 μm.

DOE models of vascular network formation

Data from the initial experiments described above were used as input in DOE models used to predict MSCs outgrowth area, EC network area, vessel length, and vessel diameter. Two DOE statistical design models were created to directly compare the traditional experimental design with a refined DOE approach. The traditional experimental design was modeled using a three-level full factorial design meaning that all experimental space would be explored. A second model was created using a d-optimal quadratic design using only 17 experimental conditions recommended from the predictive models (Table 2). Both the full factorial design and d-optimal design models are sampled from the same design space as shown in Appendix Figure A3. All data for 1 mg/mL were excluded from the model due to inconsistent EC outgrowth.

Summary statistics were used to evaluate the fit of each model and design (Appendix Fig. A4). R², the goodness of fit, was >0.8 for all responses in the full design and >0.6 for all responses in the minimal design. Q², the “goodness of prediction,” was >0.8 for all responses in the full design and >0.5 for all responses in the d-optimal design, indicating that the models are good and significant predictors. Model validity values <0.25 indicate statistically significant model problems. Model validity was −0.2 for all responses in the full factorial design (Appendix Fig. A4A) and >0.35 for all responses in the d-optimal design (Appendix Fig. A4B). Reproducibility, the variation of the replicates compared to overall variability, should be >0.5. All responses in both designs had reproducibility values >0.9 (Fig. A4).

Three-dimensional surface plots were created to visualize the full factorial and d-optimal DOE models. Both models displayed similar trends of MSCs outgrowth (Appendix Fig. A5) and EC network area (Appendix Fig. A6). However, in the case of the EC network area, the d-optimal model did not show any dependence on fibrin concentration (Appendix Fig. A6). The full factorial model predicted a maximum MSCs outgrowth area of 73.3 mm² at 85% MSCs in 1 mg/mL fibrin, and the minimal design predicting a maximum MSCs outgrowth area of 65 mm² at 85% MSCs in 1 mg/mL fibrin. The full factorial model predicted a maximum ECs network area of 2.56 mm² at 52% MSCs, 1 mg/mL fibrin. The d-optimal model predicted a maximum EC network area of 1.66 mm² at 51% MSC regardless of fibrin concentration. The two models' predictions for vessel length and diameter varied significantly (Appendix Figs. A7 and A8). Both DOE models predicted the same trends and maximum EC network area with ∼50% MSC spheroids, while MSCs outgrowth was maximized with ∼85% MSC spheroids.

The predictive capacity of DOE models

The accuracy of the DOE model predictions was investigated by the evaluation of intermediate spheroid culture conditions. Coculture spheroids containing 40%, 60%, and 70% MSC were cultured in 2 mg/mL fibrin for 3 weeks. The EC network area predicted by the full factorial and d-optimal design models under these conditions were compared to the experimental results (Tables 3 and 4).

Table 3.

Comparison Between Full Factorial Design of Experiments Model and Experimental Values

% MSC	Predicted area (mm2)	Experimental area (mm2)	Difference (%)
40	2.15	2.07 ± 0.35	3.7
60	2.25	2.18 ± 1.20	3.1
70	1.63	1.06 ± 0.26	35

Table 4.

Comparison Between d-Optimal Design of Experiments Model and Experimental Values

% MSC	Predicted area (mm2)	Experimental area (mm2)	Difference (%)
40	1.59	2.07 ± 0.35	30
60	1.61	2.18 ± 1.20	35
70	1.45	1.06 ± 0.26	27

Both designs accurately predicted the trend of EC growth, however, the full factorial design predicted values that were closer to experimental values. These data suggest that DOE could be a potential tool for the optimization of experimental parameters while minimizing time and costs.

In vitro evaluation of prevascularized fibrin units

The optimized culture conditions were utilized to develop the prevascularized components of composite fibrin/HA hydrogels, as depicted in Figure 1. A fibrin concentration of 3 mg/mL was chosen to maximize cellular outgrowth and structural integrity. Based on in vitro studies and the DOE model, to balance the outgrowth of MSCs and ECs, coculture spheroids composed of 40% iPSCs-ECs and 60% MSCs were selected for in vivo evaluation. Two spheroids were encapsulated in each fibrin unit to maximize the vascularized volume. Coculture spheroids with 40% iPSCs-ECs and 60% MSCs were suspended in 3 mg/mL fibrin gel (two spheroids per gel) and cultured over 3 weeks. Cellular outgrowth was observed with brightfield and confocal microscopy (Fig. 3). MSC outgrowth from the spheroids covered almost the entire area of the well (Fig. 3A). Immunofluorescent staining for CD31 showed that ECs sprouting from each spheroid formed interconnected networks and spanned the full diameter of the well (Fig. 3B, C). Immunofluorescent staining for the MSC marker CD90 showed MSCs surrounding the EC vessel-like structures (Fig. 3D–F).

FIG. 3.

MSC and iPSC-ECs outgrowth in multispheroid scaffolds. (A) MSC outgrowth is visible throughout the entire scaffold. (B) Brightfield image of MSC outgrowth overlaid with a mosaic confocal image of EC outgrowth shows scale of EC outgrowth. Vessel-like structures are visible as dark shadows in brightfield image and red sprouts in confocal image. (C) Connection of sprouts from two spheroids. (D–F) Confocal images of immunofluorescent staining for CD90 [MSCs, (D)] and CD31 [ECs, (E)] and overlay (F) show MSCs surrounding EC sprouts. Scale bars represent 500 μm in (A, B) and represent 200 μm in (C–F). Color images are available online.

Composite fibrin/HA scaffolds

For the in vivo study, composite scaffolds were prepared by encapsulating prevascularized fibrin units (described in section “In vitro evaluation of prevascularized fibrin units”) in HA-loaded fibrin gel (depicted in Fig. 1). SEM was used to visualize and compare the structure of fibrin and HA–fibrin composite scaffolds. Fibrin-only scaffolds had an open, porous structure (Fig. 4A). HA-loaded areas of the composite scaffolds had HA particles adhered to the fibrin sheets and around the pores (Fig. 4B, C). High magnification showed spherical HA nanoparticles along with the fibrin sheets (Fig. 4D).

FIG. 4.

Scanning electron microscopy images of fibrin scaffolds with and without HA. (A) Porous sheet-like structure visible in fibrin-only scaffolds. (B, C) Clusters of HA visible among fibrin pores in HA–fibrin composite scaffolds. (D) Microstructure of HA particles. Scale bars represent 100 μm in (A–C) and 10 μm in (D).

Cranial defect study

Prevascularized composite HA/fibrin scaffolds were implanted in a rodent critical-sized cranial defect model to evaluate the effect of preformed vascular networks on bone regeneration. Two groups were studied: (i) composite scaffolds with prevascularized fibrin units and (ii) composite scaffolds with MSC-only networks. Samples were harvested at week 1 to assess the survivability and functionality of implanted EC networks. Samples were harvested at week 8 to evaluate vascularization and bone regeneration. The fibrin scaffolds were visible in samples harvested at week 1, but not in samples that were harvested at 8 weeks. At 8 weeks, all defects were filled with tissue, and there were no observable differences in gross appearance between treatment groups.

Vascularization

Blood vessel density

Vascularization at the defect site was evaluated through immunohistochemical staining for CD31 and quantifying mean blood vessel density (number of CD31-positive vessels per mm²). Immunohistochemical CD31 staining images showed blood vessels spanned throughout the entire defect area (Fig. 5A–C). The mean blood vessel density in samples containing prevascularized scaffolds was 187 ± 34 vessels/mm² at week 1 and 156 ± 18 vessels/mm² at week 8. This decrease over time was not statistically significant. After 8 weeks of implantation, prevascularized scaffolds demonstrated significantly higher (p = 0.025) mean blood vessel density than MSC-only scaffolds (103 ± 10 vessels/mm²) (Fig. 5D).

FIG. 5.

Vascularization in cranial defect study. Immunohistochemical CD31 staining of (A) prevascularized scaffolds at 1 week harvest, (B) prevascularized scaffolds harvested at 8 weeks, and (C) MSC-only scaffolds harvested after 8 weeks. (D) Mean blood vessel density (number of CD31-positive structures divided by tissue area) quantified from images of CD31 stain. Prevascularized scaffolds had significantly higher blood vessel density than MSC-only scaffolds after 8 weeks of implantation. *p < 0.05. Scale bar in all images represents 50 μm.

Survival and function of implanted human iPSC-EC networks

Survival and function of the implanted human iPSC-ECs networks were assessed through immunohistochemical (Fig. 6) and immunofluorescent (Fig. 7) staining for human-specific CD31. At 1 week following implantation, staining for human-specific CD31 indicated the presence of implanted vessels within the defect area (Fig. 6A). Furthermore, erythrocytes were visible in the vessel lumen suggesting that the implanted vessels anastomosed with the host vasculature. Implanted vessels were also visible after 8 weeks following implantation (Fig. 6B). At 8 weeks, the vessels were larger than those present at week 1. H&E staining showed spheroids in areas close to neobone regeneration (Fig. 6C, indicated by asterisks). These spheroid structures were confirmed to contain human vessels in serial tissue sections stained for human-specific CD31 (Fig. 6C).

FIG. 6.

Presence of human iPSC-EC vessels after 1 week and 8 weeks of cranial defect implantation. Staining for human-specific CD31 showed presence of implanted iPSC-EC networks after 1 week (A) and 8 weeks (B) of implantation. Blue arrows indicate blood vessels and red arrows indicate spheroid structure. [(B), inset] Lack of positive stain in vessels in tissue outside the defect area (black arrows) indicates antibody specificity. (C) H&E stain of same sample shows spheroid and vessel structures in close proximity to areas of new mature and immature bone formation (black asterisks). Scale bars represent 50 μm in (A, B) and 100 μm in (C). H&E, hematoxylin and eosin. Color images are available online.

FIG. 7.

Human iPSC-EC vessels were perfused following cranial defect implantation. Confocal microscope images of samples receiving prevascularized scaffolds, stained for human-specific CD31 [green, (A)] and perfused with conjugated lectin [red, (B)]. Merged images (C) show colocalization of hCD31 and perfused lectin (white arrows), indicating survival and perfusion of implanted human vessels after 8 weeks of implantation. Scale bar in all images represents 50 μm. Color images are available online.

Tissue sections were stained for human-specific CD31 after perfusion of the vasculature with isolectin (Fig. 7) to determine whether the implanted human vessels were functional. The prevascularized scaffolds stained positive for human-specific CD31. Colocalization of human CD31 with isolectin staining indicates the vessels that were connected to the host vasculature at the time of harvest. Colocalization of perfused isolectin and hCD31 indicated functional implanted vessels after 8 weeks (Fig. 7C, white arrow).

MSCs have the capability to differentiate into ECs.⁴⁰ A sample with MSC-only scaffolds harvested after 8 weeks was stained with human-specific CD31. Human-specific CD31-positive structures were not visible in samples containing MSC-only scaffold (data not shown). Based on these results, it is hypothesized that the human CD31-positive structures in the prevascularized sampels are iPSC-ECs. However, it is possible that some of the CD31-positive cells resulted from the differentiation of MSCs to ECs.

μCT analysis of bone regeneration

μCT imaging was performed on explanted samples to determine bone regeneration and mineralization within the defect site. Three-dimensional reconstruction of μCT images indicated the presence of HA in scaffolds following 1 week of implantation (Fig. 8A). However, no HA was present in the defect area after 8 weeks in a prevascularized sample (Fig. 8E) or MSC-only sample (Fig. 8F). This resorption of HA suggests cellular remodeling at the defect site. Three ROIs with diameters of 7, 8, and 9 mm were selected for bone mineral analysis at the defect site (Fig. 8G). Prevascularized scaffolds demonstrated a slightly higher ratio of bone volume to tissue volume (Fig. 8I), however, this increase was not significant. In all ROIs, prevascularized scaffolds demonstrated a significant increase in bone mineral density compared to MSC-only scaffolds (Fig. 8H).

FIG. 8.

Representative 2D images and 3D reconstructions of bone regeneration visualized with μCT. (A–C) 3D reconstructions of μCT images. Host bone is shown as blue and bone within the defect area is shown as purple (all pseudocolor assigned by location and mineral density). (D–F) Maximum intensity projection in 2D. (A, D) The implanted HA is visible in the samples harvested at 1 week (prevascularized scaffold). (B, E) Prevascularized sample harvested at 8 weeks (C, F) MSC-only sample harvested at 8 weeks. Bone regeneration quantified from μCT images. (G) Schematic showing 7, 8, and 9 mm ROI compared to sample explant. (H) Bone mineral density measured within 7 mm (dark blue), 8 mm (medium blue), and 9 mm (aqua) ROIs. The MSCs-only group had significantly less mineral density than prevascularized samples at week 1 and 8 in all ROIs. (I) Bone volume as a percent of tissue (ROI) volume in 8 mm ROI. In all graphs, *p < 0.05. 2D, two-dimensional; 3D, three-dimensional; ROI, region of interest; μCT, microcomputed tomography. Color images are available online.

Histological analysis of the bone formation

To assess bone formation, harvested samples were decalcified and prepared for histological analysis. H&E and Masson's Trichrome staining were performed to examine tissue structure and bone regeneration (Fig. 9). The defect site was filled with soft tissue but lacked any bone regeneration at week 1 following implantation. Histology images at week 8 showed neobone formation at the periphery of the defect for both prevascularized and MSC-only scaffolds. Highly organized collagenous tissue and immature bone was observed throughout the span of the defect in prevascularized scaffolds (Fig. 9A, B). The amount of bone regeneration was quantified from histology images based on tissue morphology and reported as a percent of total tissue area. Prevascularized scaffolds resulted in a higher bone formation per tissue area of 40 ± 10% in comparison with MSCs-only scaffolds, which had the bone formation of 35 ± 9%.

FIG. 9.

Histology images showing bone and tissue regeneration within cranial defects. (A, B) Prevascularized samples harvested at 8 weeks stained for (A) H&E and (B) Masson's Trichrome show regions of collagenous tissue formation and bone regeneration. (C, D) MSC-only samples harvested at 8 weeks stained for (C) H&E and (D) Masson's Trichrome. Defect area is marked with dotted black lines. Areas of mature bone formation are denoted with asterisks and immature bone regeneration is denoted with arrows. Scale bars in all images represent 500 μm. Color images are available online.

Discussion

The clinical impact of bone tissue engineering constructs for regenerative medicine largely depends on postimplantation vascularization. Several strategies have been developed to promote and evaluate vascularization following implantation.^41–43 Prevascularized scaffolds have been shown to anastomose with host vasculature resulting in improved postimplantation vascularization and bone regeneration.^17,44 The objective of this study was to develop a prevascularized bone tissue construct using a cocultured system of clinically relevant cell types aided with an osteogenic component to improve vascularization and bone regeneration in the cranial defect model. In this study, we presented that prevascularized fibrin/HA scaffolds with cocultured iPSC-ECs and MSCs improved postimplantation vascularization. Implanted vascular networks were present and functional after 8 weeks, which may have important implications for clinical translation.

The coculture system of MSCs and ECs has previously demonstrated stable vascular network formation.⁴⁵ However, the application of iPSC-ECs in the present study examines a potential alternative source of ECs. iPSCs are obtained by reprogramming somatic cells through forced expression of specific transcription factors.⁴⁶ Theoretically, all types of cells can be induced to generate iPSCs and can be obtained through minimally invasive procedures such as shave or punch biopsies with minimal ethical concerns. The major advantage of iPSCs is their genetic and immunohistocompatibility match, which reduces the risk of infectious diseases and immune response.⁴⁷ This autologous cell source has been investigated for disease modeling,⁴⁸ patient-specific tissue regeneration,⁴⁹ and targeted drug discovery.⁵⁰ EC derived from iPSCs can be used for developing personalized medicine or patient-specific engineered vascularized tissues.⁵¹ Since autologous iPSC-ECs can be obtained in sufficient quantity without any ethical concerns, the iPSC-EC/MSC coculture system presents a feasible vascularization technique. However, a major challenge with large scale production of iPSC-ECs is the instability of endothelial phenotype and limited cell proliferation. This could be addressed by using iPSC-ECs that overexpress Sirtuin1, which has recently been shown to maintain EC phenotype and improve cell proliferation.⁵² In addition, iPSC-ECs may exhibit differences in vasculogenic ability based on the donor cell type.⁵³ These potential challenges must be considered before the clinical translation of iPSC-EC-based strategies.

Tissue engineering strategies often require the optimization of a large number of variables.⁵⁴ For example, in this study, the extent of vascularization depends on cell ratio, fibrin concentration, and culture time among other things. Limitations in time and cost generally result in, at best, research identifying local optima due to the inability to sample the full experimental space. DOE can be an effective tool to reduce the number of experiments required by sampling the experimental space in a systematic way.⁵⁵ DOE provides an advantage over the traditional vary-one-factor-at-a-time experimentation approach, due to the capability to alter multiple variables at a time.⁵⁶ We evaluated DOE in both a full experimental space and a smaller sample for its predicting capability.^57,58 Both designs accurately predicted the trend of EC network growth, however, the full design's predicted values were closer to experimental values. The ability to identify optimal conditions using less data indicates the potential strength of DOE. While this article provides only a proof-of-concept of the potential of statistical DOE, the results support further exploration of these tools in tissue engineering applications.

Previously, cranial defect studies performed by our group demonstrated limited bone formation in the presence of a preformed vasculature formed from HUVEC/MSC spheroids.¹⁷ To address this issue, we incorporated HA in the scaffold in this study. HA shares the chemical properties of inorganic components of the bone matrix and therefore the HA nanoparticles were added to mimic natural bone composition and geometry. However, we found that an HA concentration above 1.5 mg/mL significantly reduced vessel formation in vitro hindering the possibility of prevascularization (Appendix Fig. A9). Our finding agrees well with previous studies that demonstrated inhibition of vascular network formation in the presence of a high concentration of HA.⁵⁹ To address this issue, prevascularized fibrin units were formed first and then combined with HA-loaded fibrin to form a composite vascularized and mineralized system. The separate addition of osteogenic units allowed for a higher concentration of HA (50 mg/mL) without compromising the formation of vascular networks.

Following implantation of the composite scaffolds, the human vessels survived and were functional following 8 weeks in vivo. The overall vessel density decreased from 1 to 8 weeks, but this is to be expected with a wound healing response. An increase in vessel diameter from week 1 to 8 is consistent with vascular remodeling. Our results indicate a significant increase in vessel density at the defect sites treated with a prevascularized scaffold versus those with MSC-only scaffolds after 8 weeks of implantation. This result combined with the evidence of functional human vessels at 8 weeks suggests that the result is not due to the release of paracrine factors alone. It is commonly accepted that implanted vasculature serves as a temporary framework or bridge to promote natural healing and remodeling processes. However, in vivo retention of implanted ECs up to a year has been reported.¹⁵ Our study showed that the implanted vessels were stable for up to 8 weeks. If the prevascularized structures are long-lasting, additional screening measures may need to be taken to ensure that the engineered tissues are safe for long-term implantation.

Despite enhanced vessel density in the samples with prevascularized scaffolds compared to scaffolds with MSCs only, we did not observe a significant increase in bone formation. The prevascularized scaffolds may have improved the maturation of newly formed bone as suggested by the significant increase in bone mineral density. In flat bones such as in the skull, the mandible, and the clavicles, bone development occurs through direct conversion of mesenchymal tissue into bone, also known as intramembrane ossification.² Although the role of vascularization in endochondral ossification is well explored, its role in intramembrane ossification is not established well and is still speculative.^2,60 Therefore, further studies are necessary to determine the relationship of vascularization and intramembrane ossification. Histological assessment of scaffolds showed the extensive formation of dense, organized tissue spanning the entire defect length in both prevascularized and MSC-only scaffolds, which aligns well with results from our previous cranial defect study.¹⁷ This collagenous tissue may indicate the formation of fibrous Interzone, an initial phase of bone formation, which is often seen in distraction osteogenesis, a model of intramembranous ossification.^2,61,62 Vascular in-growth and osteoid along the fibrous collagen bundles form the zone of microcolumn. The bone formation begins at the vascular sinuses. Microcolumn of osteoid and bone formation grow toward each other and subsequently fills the fibrous Interzone. More regions of mature mineralized bone at the center of the defect may suggest the beginning of bone formation. One of the reasons for limited bone formation observed could be the implantation time, as a longer implantation time may be required to generate significant neobone formation.⁶³ Alternatively, administration of osteogenic cytokines such as bone morphogenetic protein-2, transforming growth factor-β, basic fibroblast growth factors may promote bone formation in the cranial defect model.^64,65

Conclusion

This study has demonstrated the formation of vascular networks from iPSC-EC/MSC coculture spheroids encapsulated in fibrin scaffolds. Dual-phase prevascularized fibrin and HA-fibrin composite scaffolds were developed for implantation in a critical-sized cranial defect model. The implanted iPSC-EC networks survived and anastomosed with host vasculature within 1 week of implantation and persisted throughout 8 weeks of implantation. Prevascularized scaffolds increased bone mineral density and vascularization compared to scaffolds with MSC-only networks. Bone regeneration, however, was not statistically significant, which can be attributed to implantation time or the complex interaction of vasculature and intramembrane ossification. This work is the first to demonstrate implantation and survival of iPSC-EC vascular networks in a bone defect model, showing promise for these cells for future use in improving postimplantation vascularization and other tissue engineering strategies.

Appendix

APPENDIX FIG. A1.

MSC and iPSC-EC outgrowth area quantified at weeks 1 and 2. MSC outgrowth varies with cell ratio and fibrinogen concentration after (A) 1 week and (B) 2 weeks of culture. EC outgrowth varies with cell ratio and fibrin concentration after (C) 1 week and (D) 2 weeks of culture. In all graphs, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. EC, endothelial cell; iPSC, induced pluripotent stem cell; MSC, mesenchymal stem cell.

APPENDIX FIG. A2.

MSC/iPSC-EC network parameters quantified after 3 weeks of culture. (A) MSC outgrowth area was quantified from brightfield images and varied with MSC/EC ratio and fibrin concentration. iPSC-EC network area (B) and the length (C) and diameter (D) of individual vessel-like structures were quantified from two-dimensional projections of three-dimensional z-stacks of confocal images. In all graphs, *p < 0.05, **p < 0.01, ****p < 0.0001.

APPENDIX FIG. A3.

Design spaces for full factorial and d-optimal designs. (A) Design space for full factorial (full) model with 135 experiments. (B) Design space for d-optimal (minimal) model with 17 experiments defined by the software.

APPENDIX FIG. A4.

DOE model summaries of fit. (A) Summary of fit for the four responses in the full factorial design. (B) Summary of fit for the four responses in the d-optimal (minimal) design. In all charts, R² (green) is the model fit, Q² (blue) estimates the future prediction precision, model validity (yellow) indicates various model problems, and reproducibility (aqua) describes the variation of the replicates compared to overall variability. DOE, design of experiments.

APPENDIX FIG. A5.

DOE model prediction for MSC outgrowth area at 3 weeks. The full factorial design is shown at the left and the minimal d-optimal design is shown at the right.

APPENDIX FIG. A6.

DOE model prediction for EC network area at 3 weeks. The full factorial design is shown at the left and the minimal d-optimal design is shown at the right.

APPENDIX FIG. A7.

DOE model prediction for vessel length at 3 weeks. The full factorial design is shown at the left and the minimal d-optimal design is shown at the right.

APPENDIX FIG. A8.

DOE model prediction for vessel diameter at 3 weeks. The full factorial design is shown at the left and the minimal d-optimal design is shown at the right.

APPENDIX FIG. A9.

HUVEC/MSC network formation in HA-loaded fibrin hydrogels. (A) Confocal microscopy image of CD31 staining shows formation of HUVEC networks. (B) Quantification of EC network area showed HA concentrations above 1.5 mg/mL significantly reduced EC network formation. *p < 0.05, **p < 0.01 against HA 0 mg/mL group. HA, hydroxyapatite; HUVEC, human umbilical vein endothelial cell.

Disclosure Statement

No competing financial interests exist.

Funding Information

This work was supported, in part, by funding from the National Institutes of Health (grants R01AR061460, 5R01EB020604), the National Science Foundation (CBET-1263994, IIS-1125412, EEC-1461S215), and the United States Department of Veterans Affairs (5 I01 BX000418–06).