Spatiotemporal Blood Vessel Specification at the Osteogenesis and Angiogenesis Interface of Biomimetic Nanofiber-Enabled Bone Tissue Engineering

Abstract

While extensive research has demonstrated an interdependent role of osteogenesis and angiogenesis in bone tissue engineering, little is known about how functional blood vessel networks are organized to initiate and facilitate bone tissue regeneration. Building upon the success of a biomimetic composite nanofibrous construct capable of supporting donor progenitor cell-dependent regeneration, we examined the angiogenic response and spatiotemporal blood vessel specification at the osteogenesis and angiogenesis interface of cranial bone defect repair utilizing high resolution multiphoton laser scanning microscopy (MPLSM) in conjunction with intravital imaging. We demonstrate here that the regenerative vasculature can be specified as arterial and venous capillary vessels based upon endothelial surface markers of CD31 and Endomucin (EMCN), with CD31⁺EMCN⁻ vessels exhibiting higher flowrate and higher oxygen tension (pO₂) than CD31⁺EMCN⁺ vessels. The donor osteoblast clusters are uniquely coupled to the sprouting CD31⁺EMCN⁺ vessels connecting to CD31⁺EMCN⁻ vessels. Further analyses reveal differential vascular response and vessel type distribution in healing and non-healing defects, associated with changes of gene sets that control sprouting and morphogenesis of blood vessels. Collectively, our study highlights the key role of spatiotemporal vessel type distribution in bone tissue engineering, offering new insights for devising more effective vascularization strategies for bone tissue engineering.

Graphical Abstract

1. INTRODUCTION

Repair and reconstruction of bone loss due to trauma and infection remains a significant clinical challenge. Worldwide, autografts or allografts are used in approximately 3 million orthopaedic procedures annually [1-3], of which 6% are craniomaxillofacial in nature [4-6]. Bone tissue engineering has been viewed as the ultimate solution for replacing bone autograft in repair of bone defects. However, the long-term success of bone tissue engineering is impeded by inadequate vascularization of the engineered construct [7-9]. The current lack of progress in vascularization of tissue-engineered scaffold is attributed to our incomplete understanding of angiogenesis and vascular beds during bone repair and regeneration. Better insights into the events mediating blood vessel network formation, specification, and remodeling will improve our understanding of skeletal repair and further clear the path to new therapeutic opportunities.

A functional blood vessel network consists of arteries, veins and a capillary interface that connects arterial and venous vessels for proper vascular perfusion. During tissue repair, a functional vascular bed is established via sprouting, maturation, and a series of structural adaptation of angiogenesis-derived vessels to form an effective hierarchical perfusion circuit that embodies arterial and venous identities. In addition, endothelial cells (EC) are highly heterogeneous at tissue and organ level and can be further specified according to the functional need of the local microenvironment, namely metabolic demands, hemodynamic forces, oxygen levels, and interactions with neighboring smooth muscle or mural cells [10-12]. The high degree of heterogeneity of capillary vessels at the organ and tissue level is similarly recapitulated during tissue repair and regeneration, with added complexity attributed to injury, inflammation and immune response at a healing site.

The specification of blood vessels at the capillary level during bone tissue engineering has not been previously described, however, a series of recent studies have linked bone formation to specific subsets of vessels in bone development and aging at the metaphysis of long bone. Defined by the expression of CD31 and EMCN, type H (CD31^highEMCN^high) and type E (CD31^highEMCN^low) vessels, as opposed to type L (CD31^lowEMCN^high) vessels, are shown coupling to OSX⁺ osteoblasts at the long bone metaphysis [13-15]. Both type E and type H vessels have been shown to directly connect to arterioles, exhibit high flow rate, and be regulated by molecular pathways that control arterial EC specification during development, e.g., hypoxia inducible factor (HIF), Notch and BMP pathways. A series of recent studies demonstrate a precise regulation of type H vessel that involves complex molecular signals orchestrated from osteoblasts, osteoclasts, as well as endothelium at the site of trabecular bone formation [16-18]. These studies suggest that the specification of type H vessels unique to bone tissue could play an important role in bone regeneration. In support of this novel concept, vascular control of bone formation has been demonstrated in a recent study, in which Slit3, a key family of genes that regulate nerve patterning and vascular sprouting, is shown to direct type H vessel formation and bone regeneration [19].

The overall goal of our current study was to define osteogenesis-dependent vessel specification at the osteogenesis and angiogenesis interface of bone tissue engineering. To ensure robust bone regeneration, we utilized a previously described polycaprolactone, collagen, and nano-hydroxyapatite biomimetic composite nanofiber sheet, which promotes a donor bone marrow stromal cell (BMSC)-dependent bone formation in repair of a segmental defect [20]. Our study was motivated by the clinical need to develop effective therapeutics, as well as the importance of addressing a critical knowledge gap in our understanding of the functional vascular bed in bone regeneration. Utilizing high resolution multiphoton laser scanning microscopy (MPLSM) in conjunction with a previously described cranial bone defect window chamber model, which permits high resolution real-time analyses of vascular function at the osteogenesis and angiogenesis interface at the site of cranial bone defect repair [21, 22], we examined the regenerative vessel specification and its spatiotemporal correlation with bone formation during nanofiber-mediated bone tissue engineering. Our data demonstrate an important role of spatiotemporal vessel specification during repair and regeneration, and further uncover differential vascular responses and angiogenic gene sets associated with the specification of blood vessel types in healing and non-healing cranial bone defects.

2. MATERIALS AND METHODS

2.1. Mouse strains.

The Col 1 (2.3) GFP transgenic mice, which specifically label mature osteoblasts with GFP [23], and the immunodeficient NOD.CB17-Prkdc^scid/J (NOD/SCID) mice, which are deficient in functional T cells and B cells, were purchased from the Jackson Laboratory (Bar Harbor, Maine). The Osterix-RFP^Cherry reporter mice [24], which label all osteoblastic lineages with cherry red fluorescence protein (RFP^cherry) were kindly provided by Dr. Peter Maye at the University of Connecticut Heath Science Center. The immunodeficient NOD/SCID;OSX-REP^cherry mice were established by backcrossing OSX-RFP^cherry mice with NOD/SCID mice for more than eight generations. In these mice, RFP could only be detected in mature osteoblasts and osteocytes in bone tissue. The immunodeficiency of NOD/SCID-OSX-RFP^cherry mice was confirmed by the lack of T and B cells in peripheral blood via fluorescence-activated cell sorting (FACS).

All in vivo experiments were performed using adult 8–12 week old animals housed in pathogen-free, temperature and humidity controlled facilities with a 12-hour day-night cycle in the vivarium at the University of Rochester Medical Center. All cages contained wood shavings, bedding and a cardboard tube for environmental enrichment. All experimental procedures were reviewed and approved by the University Committee on Animal Resources. General anesthesia, and analgesia procedures were performed based on the mouse formulary provided by the University Committee on Animal Resources at the University of Rochester. The animals’ health status was monitored throughout the experiments by experienced veterinarians according to the Guide for the Care and Use of Laboratory Animals outlined by the National Institute of Health.

2.2. Fabrication of PCL/Collagen/HAp composite nanofibrous mesh.

Polycaprolactone (8% w/v) (PCL, SigmaAldrich, St. Louis, MO) and Collagen type I (8% w/v) (Elastin Products, Owensville, MI) solution were prepared by dissolving in 1,1,1,3,3,3,-isopropanal (HFIP, Oakwood Products, New Orleans, Louisiana) respectively and then mixed at a volume ratio of 3:1. Nano hydroxyapatite (HAp, M K Impex Corp. Missisauga, ON, Canada) was suspended in HFIP (8% w/v) by stirring overnight and treated with ultrasound bath to achieve better suspension prior mixing with the PCL/Collagen (3:1). The ratio of PCL:Collagen:HAp is 3:1:2. Electrospinning was performed at a feeding speed of 10 μL/min by a syringe pump (Chemyx Incorporation, Houston, TX) and a 15 kV voltage power supply (Gamma High Voltage Research, Ormond Beach, FL) with a distance of 10 cm between needle and collector. Composite nanofibrous meshes were collected onto circular stainless-steel wire loops for further cell seeding and handling. The composite fiber mesh was examined by Scanning Electron Microscope (SEM) and show randomly oriented non-woven fibers with a mean diameter of ~600 nm [20].

2.3. Assembly of 3D cell/scaffold constructs.

Bone marrow cells were isolated as previously described [20, 21] from 8-12 weeks old Col 1 (2.3) GFP transgenic mice. Briefly, cells were flushed from marrow cavity by slow injection of α-MEM at one end of the bone using a sterile 21-gauge needle. The marrow suspension was dispersed gently by pipetting several times to obtain a single cell suspension. The cell suspension was further filtered through a 70μm cell strainer (Falcon) to remove debris. About 5x10⁶ freshly isolated bone marrow cells were seeded on each sterile electrospun fibrous sheet in 12-well plates and cultured in alpha-MEM media containing 15% fetal bovine serum (SigmaAldrich, St. Louis, MO) for 10 days with change of media every two days. Osteogenic differentiation media containing 50 μg/mL ascorbic acid (SigmaAldrich, St. Louis, MO), 5mM β-glycerophosphate (SigmaAldrich, St. Louis, MO), and 10% FBS in alpha-MEM was added at day 10 and cultured for an additional 11 days with medium change every two days. BMSC-seeded fiber sheets (12 layers) were stacked layer-by-layer via a custom-made metal clip to form a flexible membranous tissue construct (~200 μm in thickness).

2.4. The cranial bone defect repair model in mice.

Procedures for creating a cranial defect and mounting a glass window for imaging in mice have been previously described [21, 25]. Briefly, experimental mice were anesthetized with a mixture of Ketamine and Xylazine. Under anesthesia, hairs on the skull were removed and skin at the surgical site was sterilized with alcohol and iodide solution. A stereotaxic instrument (Stoelting Inc., Wood Dale, IL) was used to stabilize mouse head for surgery under a dissection microscope. A 2-mm in diameter full thickness defect was created in the parietal bone of mouse calvarium using a same-sized Busch inverted cone bur (Armstrong Tool & Supply Company, Livonia, MI). A circular graft of the same size was harvested from the nanofibrous construct via a biopsy punch and used to repair the defect. The skin was closed using absorbable 3-0 Ethilon sutures. Samples were harvested at the indicated time points for histology, MicroCT as well as MPLSM imaging. To perform intravital imaging during cranial defect healing, a custom-made 0.5-mm thick spacer made of poly (aryl-ether-ether-ketone) (PEEK) was glued onto the skull using cyanoacrylate glue (Loctite; Cat #45404, Düsseldorf, Germany). A glass window was mounted on top of the wound for intravital imaging as previously described [21, 25].

2.5. Multiphoton Laser Scanning Microscopy (MPLSM).

An Olympus FVMPE-RS system equipped with two 2-photon lasers: Spectra-Physics InSightX3 (680nm-1300nm) and Spectra-Physics MaiTai DeepSee Ti:Sapphire laser (690nm-1040nm), and a 25X water objective (XLPLN25XWMP2, 1.05NA), was used for high resolution imaging. With the laser tuned to 780nm, images were acquired from resonant scanners at a resolution of 512x512 pixels with the z-step size of 5 μm. The fluorescence of GFP, RFP, far-red RFP and Second Harmonic Generation (SHG) signals were collected with a 517/23-nm, a 605/25-nm, a 665/20nm, and a 390/20-nm bandpass filters (Semrock), respectively. The 2D slice viewing and 3D reconstruction of the defect were performed in Imaris (Bitplane Inc., Concord, MA) and Amira (Visage Imaging, Berlin, Germany) image analysis software. Red blood cell (RBC) velocity analyses were performed using a water-immersion objective (×25, NA 1.05) for line-scan measurements. These measurements utilized 640x640 pixel images with a pixel dwell of 10us/pixel. RBC velocity was calculated based on Radon transformation and an automated image-processing algorithm provided by MATLAB [26]. Vessel diameters were measured manually using ImageJ (National Institutes of Health).

2.6. Measurement of partial oxygen pressure (pO₂) in blood vessels via 2-photon phosphorescence lifetime microscopy (2PLM).

To examine the oxygen content in various blood vessels, 2-photon phosphorescence lifetime imaging (2PLM) was performed in cranial defect window chamber model, which allows realtime interrogation of pO₂ within each vessel at high spatial resolution [27, 28]. To perform 2PLM, 0.5μmol oxygen sensitive phosphorescence probe PtP-C343 mixed with 0.1 μmol Rhodamine Dextran (2,000,000 MW) was administered into circulation via retro-orbital injection. A two-photon microscope with a tunable Mai Tai laser (100 fs, 80 MHz; Spectra-physics, Santa Clara, CA) for excitation and a modified Olympus Fluoview 300 confocal unit was used for imaging. Excitation of PtP-C343 was performed at 900 nm. The light transmitted through the dichroic was passed through a 706/167 nm band-pass filter and directed onto a PMT (Hamamatsu R10699, Shizuoka, Japan) and a photon counting system (SR 400, Stanford Research Systems, Sunnyvale, CA) for quantification of PtP-C343 phosphorescence (λ_max 680 nm). Raw phosphorescence decay data were fit to a single exponential function after subtraction of the offset, to determine the decay time constant, τ. Using an independently measured calibration curve, τ was converted into oxygen tension (pO₂) as previously described [22]. Point scans were performed in randomly selected vessels within the defect. A mean from at least two measurements was used to determine pO₂ within each vessel. Following intravital imaging, samples were stained with CD31 antibody (Biolegend, San Diego, CA) and EMCN (Santa Cruz Biotechnology, Santa Cruz, CA) to determine the identity of the vessels.

2.7. Evaluation of cranial defect repair via histology and MicroCT.

The cranial defect samples were scanned by Viva μCT 40 system (Scanco Medical AG, Bassersdorf, Switzerland) at indicated time points post-surgery. The imaging data were anonymized and exported as DICOM files for the evaluation of graft bone formation using Amira (FEI Visualization Sciences Group, Hillsboro, OR, USA). The 2-mm circular defect region was contoured via VolumeEdit in Amira. Bone volumes within the defect were read from the Amira. For histologic analyses, samples were harvested and decalcified in 10% EDTA. Mid-sagittal frozen sections (20μm thick) were prepared via cryosectioning and stained with Hematoxylin & Eosin (H&E). High resolution digital images of the histologic sections were obtained via Olympus VS110™ Virtual Slide Scanning System (Olympus, Tokyo, Japan). Histomorphometric analyses to evaluate bone, scaffolds and fibrotic tissues in the tissue sections were performed using the VisioPharm Image Analysis Software (Hørsholm, Denmark) as previously described [29].

2.8. Immunofluorescent staining of blood vessels and microscopy.

At the end of the experiments, mice were perfused systemically with freshly made 4% paraformaldehyde via cardiac puncture followed by additional overnight fixation with the same solution. The cranial samples were fully decalcified in 10% EDTA and then treated with blocking solution containing 3% bovine albumin and 0.3% Triton X-100 in PBS overnight. All samples were incubated with CD31 (1:100 dilution, Biolegend, San Diego, CA) and EMCN (1:50 dilution, Santa Cruz Biotechnology, Santa Cruz, CA) antibodies labeled with respective fluorescent conjugates for 4-7 days at 4°C. The samples were mounted in anti-fade mounting medium and imaged via MPLSM as described above.

2.9. Quantitative and histomorphometric analyses of neovascularization at the site of cranial bone defect repair.

All samples were scanned from superior and dura side of the cranial bone for analyses. Multichannel z-series image stacks were used for visualization, 3D reconstruction, and quantitative analyses. A detailed method for quantitative analyses of blood vessels at the site of cranial defect repair has been previously described [21, 25]. Evaluation of bone accruement via SHG, an intrinsic signal from the collagen matrix in bone tissue, was performed alongside vascular analyses in Figs. 4 and 5 using image stacks obtained from multiphoton microscopy. A schematic to illustrate our analysis is shown in supplemental Fig. S2. Briefly, CD31⁺EMCN⁻ or CD31⁺EMCN⁺ vessels along with SHG and Col 1 (2.3) GFP cells were reconstructed in a 3D format using a multichannel z-series stack. To analyze nanofibrous membrane-mediated defect repair, a circular region consisting of the 2-mm defect up to 200-350 μm in depth (36 tiles of 512x512 z-series stack at 5 μm z-step) was contoured and specified as the region of interest (ROI). Based on unique SHG signals from bone tissue, bone forming and non-bone forming regions were contoured and specified within the circular region of the defect. Vessels within each region were isolated in Amira Segmentation Editor, followed by volumetric and length analyses using Autoskeleton Module combined with Filamental Editor as previously described [21, 25]. Based on the unique morphology of the vessel type, CD31⁺ (total vessels) and CD31⁺EMCN⁺ vessels were subjected for volumetric analyses whereas CD31⁺EMCN⁻ vessels were subjected for length analyses. Based on the segmented vessel network, vessel volume fractions (Vol. Fract.) (i.e., ratio of vessel volume to total volume) were read directly from the Amira. For length analyses, we used Amira’s CenterTree algorithm to generate a line-based network that was topologically equivalent to the original network. The skeleton was superimposed on the original image to assess the relative accuracy of this method. The final skeletonized vessel network was obtained by manually retracing of the skeletons using Amira’s Filamental Editor to remove false segments. Based on the skeletonized network, vessel Length Fractions (L.Fract.) (i.e., ratio of vessel length to total volume) were read from the Amira software. Quantitative and histomorphometric analyses of neovasculature were performed simultaneously with volumetric quantification of Col 1 (2.3) GFP cells and SHG using Amira Segmentation Editor and volumetric analysis protocol. Analyses were performed in a group of 4 mice, covering the entire defect regions.

2.10. Microarray analyses.

Nanofibrous constructs with or without BMSCs were used to repair the 2-mm cranial defects as described above. The implants were punched out at week 3 post-implantation and immediately immersed in liquid nitrogen for RNA isolation. The tissues were pulverized using a nitrogen-cooled mortar and pestle apparatus (Bel-Art, Scienceware, Pequannock, NJ, USA), and purified for total RNA isolation using the TRIzol system (Invitrogen, Carlsbad, CA, USA). A total of 6 samples in two groups (n=3) were prepared for analyses. Total RNA from each sample was isolated using an RNeasy Mini extraction kit. RNA quality and purity were determined using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). RNA integrity was determined by the Agilent 2100 bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). Whole mouse gene expression microarrays were performed using Clariom S mouse assay (ThermoFisher Scientific) that allows accurately measurements of gene-level expression from >20,000 well-annotated genes. The microarray assays were performed by Center for Functional Genomics at the SUNY University at Albany. The data were analyzed using Transcriptomic Analyses Console (TAC) (ThermoFisher) which provides log2 transformed expression values for statistical processing and hierarchical clustering analyses. Differentially expressed genes were selected with a p value less than 0.01, FDR (false discovery rate) p value less than 0.05 and a fold of change of more than 1.5 when comparing between groups. Heat maps were generated by TAC.

Biological processes, functional classifications and gene annotations were analyzed using Gene Set Enrichment Analyses (GSEA) with Hallmark and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway databases as well as Database for Annotation, Visualization and Integrated Discovery (DAVID) (http://david.abcc.ncifcrf.gov). To identify biological processes with significant enrichment, the distribution of genes from our data was compared with a reference annotation gene list for each gene ontology (GO) category. Normalized Enrichment Score (NES) were used for gene enrichment analysis. A nominal p value (NOM) less than 0.05 is considered significantly enriched in the annotation categories.

2.11. Quantitative PCR analyses.

Quantitative RT-PCR reaction was performed using SyberGreen (ABgene, Rochester, NY) in a RotorGene real time PCR machine (Corbett Research, Carlsbad, CA). All genes were compared to a standard β-actin control. Data were assessed quantitatively using analysis of variance, comparing relative levels of transcript expression as a function of time. All primers used for the assessment can be found in previous publications [30, 31]. Additional primer sequences can be found in supplemental data Table S1. Data are expressed as the means ± SEM.

2.12. Statistical analyses.

All data are shown as the mean ± standard error. Statistical analysis was analyzed by one-way ANOVA in GraphPad Prism (GraphPad Prism, San Diego, CA). A p value <0.05 following Bonferroni correction was considered statistically significant.

3. RESULTS

3.1. Layer-by-layer biomimetic nanofibrous construct promoted donor-progenitor cell survival while facilitating host osteoblast migration into the defect.

We have previously demonstrated an effective bone tissue engineering strategy utilizing layer-by-layer assembled electrospun nanofiber meshes seeded with bone marrow-derived stromal cells (BMSCs) as a periosteum mimetic to repair large bone defects [20]. Here we used the same multi-layered fibrous construct to repair a 2-mm cranial defect created in mice (Fig. 1A). As shown by MicroCT analyses (Fig. 1B), implantation of BMSC-seeded fibrous construct (BMSC-F) led to robust bone formation at 5-week post-surgery. In comparison, defect alone or defect treated with acellular fibrous construct (AF) showed modestly increased bone formation over 5-week period. Histologic analyses demonstrated that defect alone had only a thin layer of fibrotic tissue on top of the defect whereas defect treated with AF construct generated a significant amount of fibrotic tissue formation within the defect. In contrast, BMSC-F construct gave rise to multilayered bone formation in the defect region, further integrated with the host bone (Fig. 1C). Volumetric MicroCT analyses demonstrated a ~ 14-fold induction of bone formation in BMSC-F treated defect as compared to defect alone, and ~ 8-fold induction as compared to the defect treated with AF construct (Fig. 1D, p<0.001, n=6). Quantitative histomorphometric analyses showed a 4.2-fold increase in bone formation in BMSC-F group as compared to the controls (Fig. 1E, n=6, p<0.05). The percent fibrotic tissue area within the defect was ~2.5 fold higher in AF group than BMSC-F group (p<0.05).

Fig.1. Analyses of nanofiber-enabled cranial bone defect repair.

Schematic to show layer-by-layer enabled tissue engineering strategy to assemble multilayered tissue construct for 2-mm cranial defect repair (A). MicroCT images (B) and H&E histology (C) of the defects treated with or without BMSC-seeded fibers at week 5. Quantitative MicroCT analyses of bone volume within defects at week 5 (D). Histomorphometric analyses of bone and fibrotic tissue within the defects (E). n=6. ****p<0.0001. The representative H&E histology of BMSC-fiber-mediated healing at week 3 (F) and 5 (G). The corresponding fluorescence images to demonstrate donor Col 1 (2.3) GFP⁺ osteoblasts and host OSX-cherry⁺ osteoblasts within the defect at week 3 (H) and 5 (I). Scale bar = 500μm.

To determine the donor cell survival, differentiation, as well as the donor/host contribution to defect repair, BMSCs derived from Col 1 (2.3) GFP mice were seeded on nanofiber layers and implanted into immunodeficient NOD-OSX^RFPcherry mice [20]. Histology and fluorescent microscopy performed at week 3 and 5 post-surgery showed similar multilayered chimeric bone formation with mixed donor-derived Col 1 (2.3) GFP⁺ and host-derived OSX-Cherry⁺ osteoblasts/osteocytes embedded in new bone layers (Figs. 1F-I). OSX-cherry⁺ host osteoblasts were found along each fibrous layer and form bone on top of each layer. Comparing to samples obtained from weeks 3 and 5, more OSX-cherry⁺ host-derived osteoblasts were observed in the defect, suggesting rapid remodeling and replacing of the donor cells during repair (Figs. 1H&I).

3.2. Characterization of blood vessel types at the osteogenesis and angiogenesis interface of nanofiber-mediated repair.

Angiogenic response and blood vessel types were examined in the 2-mm cranial defects following implantation of the multi-layered nanofibrous constructs seeded with or without BMSCs derived from Col 1 (2.3) GFP transgenic mice. Samples harvested at week 3 and 5 post-surgery were scanned by MPLSM, which allows simultaneous visualization of bone via SHG, donor osteoblasts via Col 1 (2.3) GFP and blood vessels via immunostaining of CD31 and EMCN antibodies labeled with respective fluorescent conjugates. CD31, known as platelet endothelial cell adhesion molecule-1 (PECAM-1), is a pan endothelial marker that plays a major role in maintaining endothelial cell barrier function and controlling vascular permeability [32]. Global knockout of PECAM-1 leads to dramatic reduction of arterial remodeling and arteriogenesis [33]. EMCN is a type I O-glycosylated, sialic-rich glycoprotein, which is reportedly expressed by venous and capillary endothelium but not arterial endothelium [34-37]. Compared with other arterial immunofluorescence markers, namely SMA and Sca-1, which primarily labels arteriole [38, 39], CD31 and EMCN staining revealed the complete capillary interface including arterial capillary vessels of ~5μm in diameter in bone tissue.

As shown in Fig. 2, two major types of vessels were identified at the repair site: CD31⁺EMCN⁻ (red) and CD31⁺EMCN⁺ vessels (yellow or yellowish green). Both types of vessels were found in bone healing and non-bone healing regions. At week 3 post implantation, untreated defect (Figs. 2A1-3), or defect treated with AF construct (Fig. 2B1-3) showed limited defect repair as indicated by SHG signals produced by bone tissue (Figs. 2A1&B1). In the defect treated with BMSC-F construct, ample bone formation was noted in the defect (white, Fig. 2C1). Donor Col 1 (2.3) GFP⁺ osteoblasts were seen largely confined within the defect region (cyan, Fig. 2C1&2). The BMSC-F treated defects showed visibly high vessel density as compared to the other groups at week 3. High resolution zoom-in images showed that CD31⁺EMCN⁻ vessels (red) integrated and connected to CD31⁺EMCN⁺ vessels (yellow) at the capillary ends in defects only or defect treated with AF construct, forming a complete vessel network within the defects (Figs. 2A3&B3, arrows). In contrast, in the defect treated with BMSC-F construct, zoom-in images reconstructed at −150 μm below surface of the defects showed large clusters of donor Col 1 (2.3) GFP⁺ osteoblasts tightly associated with sprouting of large diameter CD31⁺EMCN⁺ vessels in the defect (Fig. 2D2). These sprouting vessels were frequently found connecting to a CD31⁺EMCN⁻ vessel that could be traced to arterioles (Fig. 2D2, arrow). In areas of newly formed bone, many CD31⁺EMCN⁻ capillary vessels were found to intertwine among large diameter CD31⁺EMCN⁺ vessels (Fig. 2D3, arrow). By week 5, visibly disorganized vessel networks were developed in control defect (Figs. 2E1-2) and defect treated with AF construct (Figs. 2F1-2). CD31⁺EMCN⁺ large diameter vessels were occasionally observed across the defects (Fig. 2E2). The overall vascularity of these defects remained similar to that of week 3. In comparison, the overall vascularity was markedly reduced in BMSC-F treated groups (Fig. 2G2 vs. C3), with more bone observed within the defects (Figs. 2G1&H1). With enhanced bone formation, large diameter type H-like CD31⁺EMCN⁺ vessels and small diameter CD31⁺EMCN⁻ vessels were found embedded in newly formed bone tissue (Fig. 2H2, arrow). Compared to week 3, fewer active Col 1 (2.3) GFP⁺ cells were identified, indicating rapid remodeling of donor Col 1 (2.3) GFP cells or decreased expression of Col 1 (2.3) GFP cells in more mature bone tissue. Representative 3D MPLSM images of three groups of defects at weeks 3 and 5, reconstructed at different depth are shown in supplemental Fig. S1.

Fig. 2. Spatiotemporal vessel specification during nanofiber-mediated repair.

Representative 3D MPLSM images of the 2-mm cranial defects at week 3 and 5, reconstructed from different combinations of channels as indicated. Defect only at week 3 (A1-3). Boxed region in A2 is shown in A3. Defect treated with AF construct at week 3 (B1-3). Boxed region in B2 is shown in B3. Defect treated with BMSC-F construct at week 3 (C1-3). Images of the same BMSC-F treated defect were reconstructed at a depth below 150μm to show vessel types associated with Col 1 (2.3) GFP⁺ cells (cyan) (D1-3). Boxed regions in D1 are shown in D2 and 3. Defect only at week 5 (E1-2). Boxed region in E1 is shown in E2. Defect treated with AF construct at week 5 (F1-2). Boxed region in F1 is shown in F2. Defect treated with BMSC-F construct at week 5 (G1-2). Images of the same BMSC-F treated defect was reconstructed at 150μm below to show vessels in newly formed bone (H1-2). Boxed region in H1 is shown in H2. Scale bar = 500μm. Arrows indicate the connections between CD31⁺EMCN⁻ (red) and CD31⁺EMCN⁺ (yellow or yellowish green) vessels. Vessels were stained with CD31 and EMCN antibodies labeled by far red and red fluorescent dye. Tiling boundaries can be seen in some images.

3.3. CD31⁺EMCN⁻ and CD31⁺EMCN⁺ vessels exhibited distinct functional properties at the cranial repair site.

Functional properties of the two types of vessels were examined during repair via intravital imaging in a cranial window chamber model (Figs. 3A&B). Partial oxygen tension (pO₂) (Fig. 3C) and RBC velocity (Fig. 3E) of vessels were examined via intravital imaging at week 3 post-surgery, followed by immunostaining of the vessels with CD31 and EMCN antibodies (Figs. 3D&F). The vessel diameters ranged from less than 5 μm to 120 μm, with the mean diameter of CD31⁺EMCN⁻ microvessels significantly smaller than CD31⁺EMCN⁺ microvessels (Fig. 3G, p<0.05). CD31⁺EMCN⁻ microvessels had a mean RBC velocity of 1.0 ± 0.55 mm/s, ~2.7 fold higher than CD31⁺EMCN⁺ vessels of the same size range (Fig. 3H). CD31⁺EMCN⁻ microvessels exhibited significantly higher pO₂ than CD31⁺EMCN⁺ vessels at the defect site (Fig. 3I, p<0.05), confirming that CD31⁺EMCN⁺ vessels were derived from venous vessels whereas CD31⁺EMCN⁻ vessels were extended from arterioles.

Fig. 3. Functional analyses of the different types of vessels at the defect repair site.

Intravital imaging was performed in a cranial window chamber model (A). Representative image of the defect perfused with vascular dye at 3-week post-transplantation (B). Boxed region in B is illustrated to show vessels with indicated pO₂ values (C). The same sample was stained with CD31 and EMCN antibodies to reveal vessel types (D). Flow rates were measured via line scans as indicated (insert) with the calculated flow rate indicated at each vessel segment (E). The same sample was stained with CD31 and EMCN antibodies to reveal vessel types (F). CD31⁺EMCN⁻ vessels shown as red, CD31⁺EMCN⁺ vessels shown as green. Measurements of vessel diameter (G), RBC velocity (H), and pO₂ (I) in CD31⁺EMCN⁻ vessels (red) and CD31⁺EMCN⁺ vessels (green). * p<0.05. n=150 vessels in 6 mice. * p<0.05. Scale bar = l00μm.

3.4. Differential angiogenic response and vessel type distribution in healing and non-healing defect repair-mediated by nanofibers.

Quantitative analyses of vascularity during cranial defect repair were performed within the circular defects at week 3 and 5 to further characterize the angiogenic response and vessel type distribution in various defects (Fig. 4, n=4 per group). Consistently, BMSC-F treated defects showed progressively enhanced bone formation as indicated by Vol. Fract. of SHG within defects, reaching −70% by week 5 post-implantation (Figs. 4C1&F1 and Fig. 4G, p<0.05). In defects treated with AF constructs, Vol. Fract. of CD31⁺(total vessel, Fig. 4H), CD31⁺EMCN⁺ vessels (depicted as green in Figs. 4 panel A to F and Fig. 4I), and Length Fract. of CD31⁺EMCN⁻ vessels (depicted as red in Figs. 4 panel A to F and Fig. 4J) were similar from week 3 to week 5 (p>0.05). Although defects only group showing lower values than defects treated with AF constructs, the difference was not significant due the variability among samples in this group (p>0.05). In defects treated with BMSC-F constructs, Vol. Fract. of CD31⁺ and CD31⁺EMCN⁺ vessels, and Length Fract. of CD31⁺EMCN⁻ vessels were significantly higher than the two controls at week 3 post-implantation (Figs. 4H&I&J, p<0.05). However, at week 5, a significant reduction of the overall vascularity was noted in defects treated with BMSC-F constructs (Figs. 4F1-4 and Figs. 4H&I&J, p<0.05). CD31⁺EMCN⁺ vessels appeared to have a greater reduction than CD31⁺EMCN⁻ vessels in BMSC treated group (Fig. 4I vs. 4J, p<0.05), suggesting a differential vascular response during repair in the defects treated with or without BMSCs.

Fig 4. Quantitative analyses of vessel types via MPLSM in cranial defects treated with or without BMSC-seeded nanofibrous constructs.

ROIs of the cranial defects were created and reconstructed to show donor GFP⁺ cells (cyan), new bone (white), CD31⁺EMCN⁺ (green) and CD31⁺EMCN⁻(red) vessels within the defects. Defect alone (A1-4, D1-4), defect with acellular fibers (B1-4, E1-4) or defect with BMSC fibers (C1-4, F1-4) at week 3 (panel A, B and C) and week 5 (panel D, E and F). Vol. Fract. of bone as evaluated via SHG (G), total CD31⁺ vessels (H), CD31⁺EMCN⁺ vessels (I), and Length Fract. of CD31⁺EMCN⁻ vessels (J) are shown. n=4 per group, * p<0.05. Scale bar = 200μm.

To further define the angiogenic response associated with bone formation, we separately reconstructed the vessels in bone forming and non-bone forming regions in defects treated with fibrous constructs with or without BMSCs (Fig. 5). In AF construct-treated group, disorganized vessels were found to persist within the non-bone forming regions of the defects over a period of 5 weeks (Figs. 5A3&B3). In BMSC-F treated group at week 3, the sprouting CD31⁺EMCN⁺ vessels were found tightly associated with clusters of donor Col 1 (2.3) GFP⁺ osteoblasts (Figs. 5C2&3) located at the center region of the defect. These sprouting CD31⁺EMCN⁺ vessels were connected to CD31⁺EMCN⁻ vessels located at the peripheral region (Figs. 5C3&4). By week 5, a significant reduction of CD31⁺EMCN⁺ vessels was observed in newly formed bone. Large diameter CD31⁺EMCN⁺ vessels were found embedded within bone tissue. The small capillary CD31⁺EMCN⁻ vessels were found extended into the center of the defect region in BMSC treated defects (Figs. 5C3 vs. D3 and C4 vs. D4). The reduction of the vascularity was also observed in non-bone forming tissues in defects treated with BMSCs (Figs. 5C5 vs. D5 vs. C6 vs. D6). Separate evaluations of the two types of vessels in bone forming and non-bone forming regions of the defect again showed higher Vol. Fract. and Length Fract. of CD31⁺EMCN⁺ and CD31⁺EMCN⁻ vessels respectively, at week 3 in BMSC-F treated defects as compared to the controls (Figs. 5E&F, p<0.05, n=4). While the Vol. Fract. and Length Fract. of the two type of vessels were reduced at week 5, the Length Fract. of CD31⁺EMCN⁻ capillary vessels in newly formed bone tissue at week 5 was higher (p<0.05), suggesting an important transition and modification of vessel type distribution during BMSC-mediated bone regeneration. The changes of the two types of vessels in non-bone forming regions in AF group showed trending down of CD31⁺EMCN⁺ capillary but upward of CD31⁺EMCN⁻ although the differences were not statistically significant.

Fig. 5. Vessel type distribution in bone-forming and non-bone-forming regions of the cranial defects.

Defects treated with fibers at week 3 (A1-4) and week 5 (B1-4) post-implantation were reconstructed to show vessels in the regions of bone forming (A1-2 and B1-2) or non-bone forming region (A3-4 and B3-4). Similarly, defects treated with BMSC-seeded fibers at week 3 (C1-6) and week 5 (D1-6) were reconstructed to show regions of bone forming (C1-4 and D1-4) and non-bone forming (C5-6 and D5-6). New bone (white), Col 1 (2.3) GFP⁺ cells (cyan), CD31⁺EMCN⁺ (green) and CD31⁺EMCN⁻ (red) vessels were reconstructed in the respective regions to show the spatial relationship of vessel types and bone/cells. Vol. Fract. Of CD31⁺EMCN⁺ (E) and Length Fract. of CD31⁺EMCN⁻ (F) in various regions of defects are shown. n=4; a, p<0.05 when comparing with or without cells in bone region at week 3; b, p<0.05 when comparing with or without cells in non-bone region at week 3; c, p<0.05 when comparing with or without cells at week 5 in bone region; d, p<0.05 when comparing between week 3 and 5 with cells in bone region; e, p<0.05 when comparing between week 3 and 5 with cells in non-bone region. Scale bar = 200μm.

3.5. Differential expression of gene sets associated with vessel morphogenesis in healing and non-healing defects.

The close interaction of Col 1 (2.3) GFP⁺ osteoblasts with sprouting CD31⁺EMCN⁺ vessels and the transition of the sprouting vessels into large diameter type H vessels within newly formed bone tissue strongly suggest that bone forming cells express paracrine factors that control the development and remodeling of the unique bone-specific vessel network in support of bone regeneration. In search of genes associated with BMSC-dependent angiogenesis and osteogenesis, we performed a microarray analysis on samples harvested from the cranial defects at weeks 3 post-implantation (Fig. 6A-C). A total of 1569 differentially expressed genes were identified with a greater than 1.5-fold change (p<0.01 and FDR p<0.05) between groups treated with fibrous scaffolds with or without BMSC (Fig. 6A). Gene Set Enrichment Analysis (GSEA) revealed significantly enriched gene sets of Gene Ontology (GO) terms of Ossification, Osteoblast Differentiation, Positive Regulation of Wnt Signaling Pathway, Response to BMP-2, Integrin Binding, and ECM Receptor Interaction (Fig. 6B). A series of genes that control vessel sprouting, filopodia formation, and vessel patterning/morphogenesis (Fig. 6C) were also identified through Database for Annotation, Visualization and Integrated Discovery (DAVID). Among them, Slit3 and several isoforms of Semaphorin gene expressions were significantly increased in BMSC-seeded implants as compared to the controls at week 3 while the expressions of VEGFa, b and c isoforms remained unchanged (Fig. 6D). These results were consistent with our imaging analyses, which demonstrated marked induction of vessel sprouting at week 3 in BMSC-F treated samples, followed by transition of blood vessel types at week 5. Consistently, Slit3 and identified Semaphorin expressions were progressively enhanced following osteogenic differentiation of BMSCs in culture (Fig. 6E), indicating an osteoblast-controlled vessel patterning and morphogenesis.

Fig. 6. Transcriptomic analyses of genes in healing and non-healing defects.

(A)The microarray volcano plot shows differentially up- and down-regulated genes in nanofibrous constructs seeded with or without BMSCs at week 3 post-implantation (n=3). (B) Enrichment plots of GO terms with indicated scores for NES, NOM, and FDR. (C) Gene expression heatmap shows an array of upregulated (top) and downregulated (bottom) angiogenesis-related genes in implants with or without BMSC treatment. (D) RT-PCR shows indicated gene expression in cranial samples. (E) RT-PCR shows progressive induction of Slit2, Slit3, and Sema3b,5a and 7a during BMSC differentiation in culture, along with the expression of osteogenic genes, namely Runx2, Sp7(OSX) and ALP. * indicates p<0.05.

4. DISCUSSION

We have previously published a biomimetic nanofiber sheet made from PCL, collagen, and nano-hydroxyapatite via electrospinning, which effectively supports donor BMSC survival and differentiation in repair and reconstruction of large segmental bone defects [20]. The biomimetic components, namely collagen and nano-sized hydroxyapatite, were selected due to their known positive effects on osteogenic cell attachment, motility, and differentiation [40-42]. The incorporation of collagen could further benefit EC differentiation and angiogenesis [43]. Here we demonstrate that this nanofiber-enabled multilayered composite construct not only promotes the survival and differentiation of donor BMSCs, but also facilitates the migration and influx of host osteoblasts into the defect, such that by week 5, the majority of the defect were occupied by host OSX⁺ bone forming cells, leading to robust and effective repair of a 2-mm cranial defect in a murine model.

Built upon the success of this construct, the focus of our current study was to establish the angiogenic response and spatiotemporal blood vessel specification at the osteogenesis and angiogenic interface utilizing high resolution MPLSM scanning in conjunction with intravital imaging. We showed that CD31 and EMCN could mark arterial and venous capillary vessels at the osteogenesis and angiogenesis interface, with CD31⁺EMCN⁻ vessels derived from arterioles exhibiting higher flow rates and higher pO₂ whereas CD31⁺EMCN⁺ vessels derived from venous vessels displaying lower flow rates and lower pO₂. The fluorescent staining of CD31 and EMCN further allowed for 3-dimensional reconstruction of the interface of arterial and venous capillary networks at the site of cranial defect repair, permitting examination of the differential dynamics and spatiotemporal distribution of the two types of blood vessels in healing and non-healing bone defects. To our knowledge, our study was the first to delineate the arterial and venous capillary vessel specification in biomaterial-mediated bone tissue repair and reconstruction. The findings from our study could contribute to a better understanding of the functional blood vessel development and morphogenesis in bone regeneration, offering new insights for developing strategies to enhance the performance of bone tissue engineered constructs.

Establishing an organized functional blood vessel network is essential for successful bone tissue repair and regeneration. The vessel types associated with bone forming cells in bone repair and regeneration have not been well defined. Utilizing high resolution MPLSM, a distinct angiogenic dynamics of the two types of vessels and a unique coupling of donor osteoblasts to CD31⁺EMCN⁺ vessel was demonstrated during nanofiber-mediated repair. At week 3 post-implantation, the sprouting CD31⁺EMCN⁺ vessels were found tightly associated with the clusters of donor GFP⁺ osteoblastic cells at the central region of the defect. These vessels were seen connected to arterial CD31⁺EMCN⁻ vessels located at the peripheral of the defect suggesting a cooperative role of arterial and venous vessels in early repair and regeneration. The sprouting of the venous vessels could be due to low pO₂ and hypoxia at the center region of the defect where donor osteoblasts accumulated. These newly formed vessels were quickly receded or remodeled as more bone tissue was deposited at the site of repair. By week 5, CD31⁺EMCN⁺ vessels were remodeled into large diameter type H like vessels with CD31⁺EMCN⁻ arterial vessels further extending into the center region of the defect (Fig. 5). The coupling of donor osteoblasts with sprouting CD31⁺EMCN⁺ vessels promotes the early restoration of blood perfusion following injury, whereas the remodeling and formation of type H vessels together with the extension of CD31⁺EMCN⁻ arterial capillaries into bone tissue enables the establishment of a functional blood vessel network capable of offering a sustained supply of fresh nutrients and oxygen to newly formed bone.

The CD31^highEMCN^high type H vessels have been well characterized in trabecular bone tissue at the metaphysis region beneath the growth plate of the long bone where active modeling and remodeling of bone tissues take place [13-15]. These vessels are considered to be “arterial capillaries” directly connected to arterioles [44]. In our current study, we were unable to define type H vessels simply based on the fluorescence intensity of CD31 and EMCN. The CD31⁺ EMCN⁺ vessels were identified in all types of defects regardless of the presence or absence of bone forming cells. However, we did observe the large diameter CD31⁺EMCN⁺ type H-like vessels within vascular channels of the newly formed bone. We believe that the type H vessels are formed upon bone regeneration in response to the signals from bone tissue and osteoblasts. The remodeling of CD31⁺EMCN⁺ vessels gave rise to type H vessels that lead to the formation of vascular channels in newly formed bone. These vessels could be further modulated according to the local bone healing microenvironment, namely hypoxia, hematopoiesis and metabolic demands of the resident cells. Our data highlight the heterogeneity of vessel subtypes and the need to identify more endothelial cell markers to define the capillary vessel types at the site of bone healing.

The intricate relationship between osteogenesis and angiogenesis has been well documented [45-47]. On one hand, coupling of osteogenesis and angiogenesis is essential for bone repair and regeneration [48]. On the other hand, excessive angiogenesis could be inhibitory to bone repair [49, 50]. While robust blood vessel ingrowth is required to facilitate bone formation, coordinated interactions between osteogenic and angiogenic cells are crucial for organized and effective repair and regeneration [8, 51]. By comparing angiogenic dynamics in different regions of bone defects, our data suggest a well-controlled angiogenic response coupled with donor osteoblasts and bone formation at the healing region of the defect, yet a persistent disorganized angiogenic response associated with the non-healing bone defect. While the vessels in the healing bone region quickly developed into type H vessels by week 5, the vessels in the non-healing defects persisted and some developed into hypertrophic venous vessels after 5 weeks following injury suggesting a differential angiogenic response and vessel type distribution in healing and non-healing bone defects.

Consistently, analyses of gene expression profiles of nanofiber implants with or without osteogenic BMSCs reveal differential gene expression in healing and non-healing bone defects at the peak of angiogenesis at week 3 post-implantation. Along with marked induction of genes associated with bone formation/skeletal development, namely SP7, RUNX2, and a myriad of bone matrix proteins, GSEA analyses revealed significant changes of a series of angiogenic genes that control sprouting and morphogenesis of blood vessels, specifically a subset of genes known to serve as attractive or repulsive guidance cues for morphogenesis of vessels and nerves. Remarkably, these genes were found progressively enhanced during osteogenic differentiation of the BMSC culture over a period of 21 days, suggesting osteoblast-controlled vessel morphogenesis during bone formation and regeneration. Among them, the Slit family proteins are known secreted repulsive axon guidance molecules that bind to Robo receptors [52-54]. Prior studies show that Slit3 regulates endothelial cell proliferation and migration [55-58]. Further study shows that Slit3 stimulates neovessel sprouting ex vivo and new blood vessel growth in vivo. The siRNA knockdown of Slit3 in mesenchymal stem cells (MSCs) leads to disorganized clustering of ECs. Knockdown of its receptor ROBO4 in ECs abolishes the generation of functional human blood vessels in an in vivo xenogenic implant [59]. In addition to its role in angiogenesis, recent studies demonstrate that Slit3 stimulates osteoblast migration and bone formation in vivo. Slit3 secreted from osteoblasts controls type H vessel formation and is further required for fracture healing in mice [19, 60]. In addition to Slit 3, several isoforms of Semaphorin and Netrin receptor UNC5B were also increased in BMSC-seeded implants. Semaphorins are known to play important roles in guiding the migration of cells and axons in the developing nervous system and have been shown to have both angiogenic and angiostatic functions [61]. Semaphorins, which signal through neuropilin 1 (Nrp1), are also shown to promote bone metabolism and bone homeostasis [62-66]. Among them, Semaphorin 3A and 4D have been shown involved in bone remodeling [67, 68]. Semaphorin 3B can be induced by 1, 25 dihydroxyvitamin D3 in osteoblasts and stimulates osteoclast differentiation [69]. Semaphorin 7A enhances osteoblast migration and osteoclast formation via increased phosphorylation of Erk1 and 2 in osteoblasts [70]. Semaphorin 5A is involved in development of normal cranial vascular system [71]. More studies are needed to define the role of this family of genes in bone regeneration. UNC5B is known to play a critical role in arterial vessel development and in blood vessel morphogenesis. UNC5B expression is vascular-specific and restricted to arterial ECs and endothelial tip cells in sprouting capillaries. The increased expression of UNC5B coincides with increased sprouting and arterial vessel extension and branching [72, 73]. These data are in line with our analyses that demonstrate increased arterial capillary vessels in BMSC seeded implants and the controlled morphogenesis of functional vessel network associated with osteoblasts and bone formation. Taken together, the altered expression of the genes that control the directional guidance cues in BMSC-seeded bone forming implants strongly suggests an important yet un-deciphered role of these genes in the establishment and morphogenesis of the unique vessel network for bone tissue formation and regeneration. Further analyses could shed lights on the mechanisms of the molecular interactions between bone and vessel forming cells during repair and regeneration.

Despite the marked differences in the expression of many osteogenic marker genes, our analyses using total RNA isolated from the defects showed no significant differences in the expression of VEGF a, b, c at week 3 in samples treated with acellular fibers or BMSC-seeded fibers. Since VEGFs can be expressed by many cells as well as inflammatory tissues at the defect, it is not surprising that the normalized expression of VEGF genes in healing and non-healing bone defects was similar at week 3. These data suggest that: 1) BMSCs may elicit an early angiogenic response that is not solely dependent upon VEGFs, and 2) the expression and effects of VEGFs from donor progenitor cells are transient and localized. To this end, the spatial transcriptomic analyses will be instrumental in establishing a more precise spatial gene expression patterns in donor osteogenic cells for a better interpretation of the role of VEGFs in BMSC-mediated cranial defect repair and reconstruction.

5. CONCLUSIONS

Vascularization and blood vessel specification during bone tissue engineering are poorly understood. The limited knowledge hinders further efforts aimed at engineering effective and functional vessel networks for enhanced repair and regeneration. A deeper understanding of the spatiotemporal regulation of angiogenesis and the interaction of tissue-engineered constructs with host vascular microenvironment is critically important for long-term success of nanomaterial-based tissue repair and reconstruction. Combining tissue engineering strategies and advanced imaging technology, our current study demonstrates a key role of vessel specification in control of bone formation in healing bone defects. Our data suggest that spatiotemporally controlled modification of vascular environment and selective integration of specific subtypes of endothelium in synergy with bone regeneration could be important steps towards achieving effective and uniform bone regeneration.