Engineering of an osteoinductive and growth factor-free injectable bone-like microgel for bone regeneration

Abstract

Bone autografts remain the gold standard for bone grafting surgeries despite having increased donor site morbidity and limited availability. BMP-loaded grafts represent another successful commercial alternative. However, the therapeutic use of recombinant growth factors has been associated with significant adverse clinical outcomes. This highlights the need to develop biomaterials that closely approximate the structure and composition of bone autografts, which are inherently osteoinductive and biologically active with embedded living cells, without the need for added supplements. Here, we developed injectable growth factor-free bone-like tissue constructs that closely approximate the cellular, structural and chemical composition of bone autografts. We demonstrate that these micro-constructs are inherently osteogenic, and demonstrate the ability to stimulate mineralized tissue formation and regenerate bone in critical-sized defects in-vivo. Furthermore, we assessed the mechanisms that allow hMSCs to be highly osteogenic in these constructs, despite the lack of osteoinductive supplements, whereby YAP nuclear localization and adenosine signaling appear to regulate osteogenic cell differentiation. Our findings represent a step towards a new class of minimally invasive, injectable, and inherently osteoinductive scaffolds, which are regenerative by virtue of their ability to mimic the tissue cellular and extracellular microenvironment, thus showing promise for clinical applications in regenerative engineering.

1. Introduction

Bone is a complex multifunctional organ that is composed of living cells and a supporting matrix of organic and inorganic materials with an intrinsic capacity of healing upon injury.^[1] Bone healing may be limited in the event of significant trauma, and may result in delayed fracture union, or complete non-union.^[2] More than 800,000 skeletal repairs, including large fractures, trauma, tumor resections, and bone augmentation procedures, are performed annually in the United States alone.^[3] Bone autografts remain the gold-standard material in treating critical-sized skeletal defects. However, the high hospitalization costs, donor-site morbidity, and limited availability of donor tissue per patient make autologous grafts a less-than-ideal material for bone regeneration.^{[2a, 4]} Other grafting materials have been introduced to address the limitations of autologous bone, however, all with limited efficacy.^[4] Currently, synthetic bone grafts fail to mimic the structure, composition, and cellular microenvironment (cells embedded in minerals) of native bone. Continuous efforts have been made to develop biologically compatible osteogenic scaffolds that mimic the advantageous properties of autologous bone. The use of therapeutic growth factors, specifically bone morphogenetic protein-2 (BMP-2), has evolved as an alternative treatment option to bone grafts to enhance bone formation in treating musculoskeletal injuries,^[5] and has become a viable option in orthopedic and dental procedures, such as spinal fusion, fracture repair, and bone augmentation. However, surgical complications of BMP-2 grafting include severe side effects such as heterotopic bone formation, osteolysis, and inflammation.^[6] These complications indicate the need for advancement in bone tissue regenerative engineering strategies ^[7] that recapitulate autograft-stimulated bone healing while minimizing adverse reactions.

During intramembranous bone formation, osteoprogenitor cells undergo osteogenic differentiation, secrete an osteoid matrix, and become progressively embedded within a matrix material that is increasingly calcified on a nano-scale level. This process of calcification is directed over time through a biomolecular process of non-classical intrafibrillar collagen biomineralization (Schematic S1A and B, Supporting Information), where proteins such as osteopontin and osteocalcin inhibit the classic nucleation and crystal growth of extracellular ions (calcium [Ca²⁺] and phosphates [PO₄³⁻]) in order to promote mineral infiltration into the intermolecular spaces of collagen fibrils.^[8] This leads to the formation of nano-scale hydroxyapatite crystallites within collagen fibrils, which confers strength and stiffness to the bone tissue, and allows it to withstand much of the cyclic loading bone endures over a lifetime. Osteoprogenitor cells sense these matrix changes and progressively change from an undifferentiated lineage to an osteoblastic phenotype, ultimately maturing into a terminal lineage that is marked by the upregulation of osteocyte-related markers, such as dentin matrix protein 1, podoplanin (PDPN), sclerostin, and others. Developing biomaterials that replicate this process of nano-scale calcification of cells embedded in a 3D matrix – like that happening in native bone –has remained a challenge in the field. ^{[7b, 9]} Conventional in vitro approaches of cell differentiation have relied on exogenous stimulation of stem cells leading to the secretion of small mineralization nodules that fail to replicate the 3D calcification of cells within the bone matrix. Similarly, composite hydrogels embedded with hydroxyapatite nanoparticles, or pre-calcified materials that are seeded with cells after calcification, also fail to replicate the bone matrix and the encapsulation of cells within a mineralized material.

We have recently developed a nano-scale materials engineering strategy to address this need, and have expended significant efforts to characterize the process of calcification of three-dimensional cell-laden collagen constructs that mimic the native bone cellular and extracellular matrix (Figure 1A, Schematic S1C, Supporting Information) with nanometer-scale precision.^{[8, 10]} The mechanistic and regenerative properties of these innovative bone-like grafts, however, have yet to be tested. Here we hypothesize that, given the ability of these scaffolds to mimic the process of cell-laden hydrogel calcification of native bone, despite their lack of growth factor supplementation, these scaffolds can regenerate critical size skeletal defects at levels that rival some of the clinical standards in the field, such as BMP2-loaded collagen. To test that, we first engineered novel cell-laden mineralized microgels as injectable scaffolds that can be delivered into a defect site minimally invasively. We then characterized the mechanisms regulating the differentiation of stem cells embedded in these mineralized constructs, and identified the effects of matrix stiffness and ion-induced osteogenic differentiation. Ultimately, we demonstrate that these scaffolds closely approximate some of the biological steps of native osteogenesis occurring in vivo, and that this can be used as a growth-factor-free alternative to bone autografts or BMP-loaded bone grafts.

Figure 1. Nano-scale mineralized injectable cell-laden collagen microgels.

A) The schematic depicts morphological changes of collagen fibrils and cells due to intrafibrillar nano-scale mineralization. The deposition of calcium phosphate leads to the formation of mineral crystallites that increases the stiffness of the constructs. These material characteristics promote osteogenic differentiation of encapsulated hMSCs and bone regenerative potential in the rodent critical-sized calvarial defect model. Photographs, magnified images, and viability assay (after 72 hrs culture) of hMSCs in non-mineralized and mineralized collagen in the form as B) monolithic and C) micro-sized constructs (900 × 900 μm) created from micro-sectioning of the monolithic constructs. D) TEM images depict the characteristic D-banding pattern of non-mineralized collagen fibrils and homogenous inter and intrafibrillar mineral apatite crystallites formation as platelets of mineralized collagen fibrils, Scale bar: 200 nm. Elemental analysis (EDAX) shows the presence of CaP in the mineralized constructs. Immunofluorescence staining images of cytoskeletal f-actin (green), focal adhesion protein vinculin (red), and nucleus (blue) of hMSCs in non-mineralized and mineralized constructs after 72 hrs culture, scale bar: 50 μm. SEM images of hMSCs embedded in non-mineralized and mineralized constructs. E) Cell viability of non-mineralized and mineralized constructs before and after micro-sectioning (n = 3). Quantification of F) spreading area (n = 3), G) cytoskeletal actin (n = 3), and H) vinculin staining of cells (n = 3) encapsulated in non-mineralized and mineralized constructs. I) Transcription profile of genes (n = 6) relevant to osteogenesis for hMSCs in the constructs cultured for 14 days in the growth medium, osteogenic induction medium (OIM), and mineralization medium for 3 days, followed by cultured in the growth medium for 11 days (Bar graphs of 68 genes are given in Figure S5, Supporting Information). Relative expressions: red (high) and blue (low). Gene expression data are shown as fold expression of target genes after normalization to the negative control, non-mineralized constructs cultured in the growth medium. Data are presented as mean ± standard error of the mean. An ordinary two-way ANOVA (cell viability) and one-way ANOVA (cell spreading area, actin, and vinculin expression) with Tukey’s multiple comparison test were used for statistical analysis. Asterisks indicate P values with statistical significances (*P < 0.05; **P < 0.01).

2. Results

2.1. Engineering injectable mineralized cell-laden collagen microgels

In order to engineer injectable mineralized microgels that approximate the key characteristics of native bone, including the presence of osteoprogenitor cells embedded in a heavily calcified collagenous matrix without osteoinductive supplements, we built upon our recently published protocol of nano-scale biomimetic mineralization of cell-laden collagen (Figure 1A and Schematic S1C, Supporting Information).^[10] A micro-sectioning technique was used to prepare and optimize micro-sized non-mineralized and mineralized constructs ranging from 100 μm to 1000 μm in size (Figure S2A and B, Supporting Information). Unlike monolithic hydrogels, these micro-sized constructs are injectable even after mineralization, and can be administered to the target defect site using a syringe needle in a minimally invasive way (Figure S2A and S6D, Supporting Information) without affecting cell viability. Furthermore, we have demonstrated that the pore spaces created between micro-sized gels upon implantation facilitate vascular invasion and nutrient diffusion during healing.^[11] Figures 1B and C show the transformation of non-mineralized constructs from a translucent appearance into an opaque matrix after the nano-scale mineralization (Figure S3, Supporting Information), wherein the hMSCs were homogeneously embedded. Importantly, transmission electron microscopy (TEM) images corroborated the apparent D-banding periodicity of the collagen fibrils, which mimics native bone ^[10] and intrafibrillar nano-scale crystallites within the mineralized collagen fibrils (Figure 1D). Of note, after 72 hrs, calcification is exclusively induced by the medium composition (Ca, P and mOPN), and not particularly influenced by cell-secreted minerals.

To test the effect of the nano-scale mineralization and micro-sectioning on the viability of the embedded cells, we performed AlamarBlue and live and dead cell assays. There was a ~3 and ~3.6-fold increase in hMSCs numbers in non-mineralized versus mineralized monolithic constructs, respectively, from days 0 to 3 (Figure S2F, Supporting Information). There were no significant differences in cell proliferation between the groups. Cell death was negligible and no significant differences in cell viability were observed between groups. During preparation of the material, the micro-sectioning step did not affect cell viability (Figure 1B, C, and E). Moreover, the mineralized constructs exhibited a compressive modulus of 150 kPa, which is nearly 216-fold higher than that of non-mineralized constructs (Figure S3I, Supporting Information), and corroborate the drastic increase in stiffness of single mineralized collagen fibers, which jump from <1 kPa to >2 GPa in 72 hrs, as we have characterized previously (Figure S3J, Supporting Information).^[10] Despite such an increase, the high concentration of collagen and water within the construct, results in bulk mechanical properties that allow it to be injected without hydrogel rupture or cell death.

To test the effect of the stiff mineralized matrix on cell behavior, staining of the actin cytoskeleton was performed. hMSCs in non-mineralized and mineralized constructs exhibited a different morphology (Figure S2G, Supporting Information). In particular, hMSCs embedded in non-mineralized gels displayed a spindle-shaped morphology with extended multiple filopodia and oblate-shaped with the polygonal dendritic structure in the mineralized constructs (Figure S2G, white arrowhead, Supporting Information). SEM images further confirmed that cells are interdigitated with the mineralized matrix, as observed in the calcified matrices of native bone (Figure 1D). The spreading areas of hMSCs embedded in non-mineralized and mineralized constructs were 1325 μm² and 966 μm², respectively (Figure 1F). These results indicated that the spreading area of hMSCs was significantly reduced when embedded in mineralized versus non-mineralized constructs. Immunofluorescence staining images showed significantly higher expression of vinculin and formation of prominent actin stress fibers in hMSCs of mineralized when compared with non-mineralized constructs (Figure 1D, G, and H, Figure S4A, Supporting Information).

In order to verify the ability of the mineralization process to induce osteogenic differentiation of stem cells in the absence of exogenous supplements, we utilized a commercially available osteogenesis gene array kit (Figure S5, Supporting Information). We compared hMSCs embedded in mineralized constructs versus hMSCs embedded in non-calcified collagen in either osteogenic induction (OIM) or basal medium, the latter which was used as a baseline for quantification of fold-increase. Figure 1I shows selected genes from a panel of 68 markers. Common osteogenic markers, such as alkaline phosphatase (ALPL), Matrix Gla-proteins (MGP), and phosphate-regulating gene with homologies to endopeptidases on the x chromosome (PHEX), were either significantly higher or comparable to OIM-treated samples. The expression of these genes confirms the ability of the calcified matrix to stimulate osteogenic differentiation of hMSCs, despite the absence of exogenous osteoinductive supplements. Since both substrate stiffening and PO₄³⁻ ion-induced signaling have long been associated with an increase in osteogenic differentiation of stem cells,^{[10, 12]} and both are an inherent phenomenon sensed by cells embedded in our scaffolds, we asked whether the presence of PO₄³⁻ ions or the rapid increase in stiffness of our mineralized constructs might be key contributors to the high osteogenic differentiation of stem cells that we found, potentially via mechanotransduction and ionic-mediated differentiation pathways.

2.2. Mechanotransduction and adenosine signaling regulate osteogenic differentiation of hMSCs embedded in mineralized hydrogels

To address the role of stiffness in mediating cell differentiation in our gels, we quantified the nuclear translocation of YAP in hMSCs encapsulated within the non-mineralized and mineralized hydrogels. Our data showed that YAP levels were significantly higher in the mineralized than in the non-mineralized or OIM cultured samples (Figure 2A). YAP localization remained cytoplasmic in cells in both the control and OIM, but was predominantly nuclear in cells of the mineralized constructs (Figure 2A). To validate the activation of YAP through RhoA in hMSCs of the mineralized constructs, we inhibited ROCK using (1R,4r)-4-((R)-1-aminoethyl)-N-(pyridin-4-yl) cyclohexanecarboxamide (Y27632).^[13] Immunofluorescence staining of YAP was significantly reduced in the Y27632-treated mineralized constructs, confirming that mineralization-induced cell differentiation is at least partially dependent on RhoA activity (Figure 2B and F). Moreover, results demonstrated that the Y27632 treatment decreased the immunofluorescence staining for RUNX2 by nearly 3.3-fold in the mineralized constructs (Figure 2C, D, G, and Figure S4B, Supporting Information), whereas no discernible effects were seen in OIM and non-mineralized gels, thus supporting the role of mechanotransduction on osteogenic differentiation (Figure 2D, G, H, and Figure S4B, Supporting Information).

Figure 2. Nano-scale mineralization-induced osteogenic differentiation via mechanotransduction and A2b adenosine receptor signaling.

A-E) Immunofluorescence staining for YAP (red), RUNX2 (green), nucleus (blue) after 7 days culture treated with and without the pharmacological inhibition of ROCK (Y-27632) and adenosine A2b receptor (PSB603) on a growth medium, OIM, and mineralization medium for 3 days followed by 4 days culture in growth medium (Scale bars: 50 μm). Non-mineralized constructs were contracted when cultured in a growth medium, and the nuclei were close to each other. F) Quantification of YAP expression (n = 7) for hMSCs cultured on a growth medium, OIM, and mineralization medium, with and without the treatment of Y-27632. G) Quantification of RUNX2 expression (n = 7) for hMSCs cultured on a growth medium, OIM, and mineralization medium, with and without the treatment of Y-27632 and PSB603. The schematic model depicts the mechanism of nano-scale mineralized constructs stimulating osteogenic differentiation via H) Rho/ROCK and I) adenosine signaling. An ordinary one-way ANOVA with Tukey’s multiple comparison test was used. Error bars are mean with s.e.m., and the significant effect of different groups on YAP and RUNX2 expression is denoted as * (p < 0.05), *** (p < 0.001), and **** (p < 0.0001), respectively.

To delineate the influence of ion-mediated cell differentiation in our mineralized constructs, we targeted a phosphate-ATP-adenosine metabolic signaling that has been shown to regulate differentiation in calcium and phosphate ceramic scaffolds, using an A2b adenosine receptor inhibitor, 8-[4-[4-(4-Chlorophenzyl)piperazide-1-sulfonyl)phenyl]]-1-propylxanthine (PSB603).^[14] The PSB603 treatment decreased the immunofluorescence staining of RUNX2 by nearly 1.4-fold in OIM constructs, and nearly 5.9 fold in the mineralized constructs (Figure 2E, G, I, and Figure S4B, Supporting Information), indicating the dependence of phosphate-dependent adenosine signaling on osteogenic differentiation predominantly in the mineralized constructs. Interestingly, the effects observed with inhibition of either adenosine or Rho-ROCK were statistically comparable in the mineralized constructs, indicating the occurrence of several mechanisms (ionic- and mechanical-dependent) of cell differentiation, due to mineralization of cells embedded in collagen. Since, numerous studies have shown the molecular mechanism for the role of YAP and adenosine signaling in bone differentiation, we have performed immunofluorescence staining-based analysis in this study.^[15] However, a future study is warranted to determine the combinatorial effect of various mechanisms involved in the osteogenic differentiation of cells in our system using molecular biology methods.

2.3. Injectable mineralized constructs promote mineralized tissue formation and long-term retention of viable hMSCs

To evaluate the regenerative potential of the mineralized injectable microgels, we used a well-established critical-sized rodent calvarial defect model (Figure 3A, Figure S6 A–D, Supporting Information).^[16] The maximum effective dose of BMP-2 (2.5 μg/defect) was chosen as a positive control as there was maximum bone formation and minimal adverse reaction as previously defined in a similar model by Pelaez et al.^{[5b, 5c]} Empty, non-grafted defects were used as negative control. All defects in the cell-laden mineralized constructs and BMP-2 treatment groups were bridged with enhanced bone formation by 12 weeks, whereas the non-mineralized constructs and empty defects had negligible bone formation, as confirmed by 3D reconstructed microcomputed tomography (μCT) images (Figure 3B and C). Longitudinal bone formation was qualitatively evaluated at weeks 0, 4, 8, and 12 (Figure S6E, Supporting Information), and the bone volume (BV in mm³) and bone bridging area (%) at 12 weeks were evaluated using μCT. The average BV (Figure 3F) measured for the cell-laden mineralized treated group (25.4 mm³) was comparable to the BMP-2 treated group (19.6 mm³). These values were significantly greater than the cell-laden non-mineralized group (6.9 mm³) and empty defects (1.7 mm³). Importantly, the average BV measured to be 0.07 mm³ (0 week) and 25.4 mm³ (12 week) for the cell-laden mineralized, indicating that the newly formed mineralized tissue (25.37 mm³) was achieved by the implanted mineralized cell-laden microgels. BMP-2 and cell-laden mineralized groups had the greatest bone bridging areas of 81.7% and 83.7% (Figure 3G), respectively. The non-mineralized constructs had 34.3%, and the empty defects had 11.6%. The cell-laden mineralized and BMP-2 groups had statistically greater bone bridging areas than the non-mineralized group and empty defects. Importantly, the average bone mineral density (BMD, Figure 3H) measured at the core of the defect for the cell-laden mineralized group (873.5 mg HA/cm³) was comparable to the BMP-2 treatment (912.3 mg HA/cm³). These values were significantly greater than the non-mineralized group (121.8 mg HA/cm³) and empty defects (0 mg HA/cm³). While bone volume, bone bridging area, and BMD (Figure 3F–H) have similar values between the BMP-2 and cell-laden mineralized groups, the surface roughness is qualitatively higher in the cell-laden mineralized group. The BMP-2 treatment resulted in a more continuous shelf of the tissue rather than a large mass of granules of bone chips, more akin to what is seen in autologous bone grafts.^{[16a, 17]} Therefore, there is room for improvement in the delivery or amount of mineralization to promote a more rapid fusion of the cell-laden mineralized microgels. Moreover, It should be noted however, that the quantified measures of BV and BMD may account for both newly formed bone, as well as the apposition of Ca and P ions available in the extracellular milieu onto the implanted mineralized scaffolds. While not possible to distinguish these sources of mineralised material via μCT, analysis of acellular mineralised constructs (Fig. S9) suggest spontaneous mineralisation may constitute a non-negligible background to analysis of bone formation. Given this background, it is difficult to draw a formal comparison of osteoinductive efficacy between the BMP-2 and cell-laden mineralized groups via μCT analysis.

Figure 3. Mineralized cell-laden collagen constructs promote bone regeneration comparable to BMP-2 in a critical calvarial defect.

A) Schematic illustration of the 8 mm critical calvarial defect model and treatment groups. B and C) Representative microCT reconstructed images showing transverse and sagittal planes view of each treatment group at week 12 post-transplantation. D) The regenerated bone regions are analyzed via SEM images. Red and blue boxes indicate the respective high magnification images of inner defect regions that show the presence of remodeling constructs, osteoid, osteocytes, blood cells (arrowheads), and blood vessels in the newly formed bone. E) Immunofluorescence staining of α-SMA for the samples treated with different groups. F) Bone volume (n = 4), G) % bone bridging area (n = 4), and H) bone mineral density (BMD, n = 4) of the newly formed bone in the defect measured by microCT at 12 weeks post-transplantation. I) A positively stained area was quantified for α-SMA (12 weeks treated, n = 4) via ImageJ. An ordinary one-way ANOVA with Tukey’s multiple comparison test was used. Error bars are mean with s.e.m., and the significant effect of different groups on bone regeneration is denoted as * (p < 0.05), ** (p < 0.01), *** (p < 0.001), and **** (p < 0.0001), respectively.

Scanning electron microscopy (SEM) was used to investigate the topography of cross-sectioned explants in cell-laden non-mineralized and mineralized constructs (Figure 3D). In the BMP-2 and cell-laden mineralized groups, bone apposition was evident primarily stemming from the periphery of the defect. The high magnification images revealed the formation of bone tissue with the presence of osteocytes within the cell-laden mineralized constructs. Both osteocytes in the cell-laden mineralized constructs and the BMP-2 group had areas of high backscatter in the surrounding lacunae, consistent with the presence of osteocyte cement lines. In higher magnification, the morphology and topography of the regenerated bone in the cell-laden mineralized samples were similar to the BMP-2 treated samples (Figure 3D). Only a thin band of fibrous tissue was seen in the empty defects; similarly, defects treated with non-mineralized constructs were primarily filled with fibrous connective tissue, as a result of collagen remodeling. Both empty defects and non-mineralized groups lacked distinct bone tissue formation. The presence of blood vessels (RBC denoted by arrowheads, Figure 3D), were notably present around the implanted cell-laden mineralized microgels. These results corroborate immunostaining data showing a positively stained area quantified for α-SMA to be significantly higher for mineralized constructs than the BMP-2, non-mineralized or empty groups (Figure 3E and I).

The newly formed bone tissue was then assessed via H&E staining. Whole-mount images highlight the native and regenerated bone tissue with high magnification images focused on the regenerative interface between existing and new tissue (Figure 4A and B). The cell-laden mineralized samples had newly formed lamellar bone circumferentially around each construct with apparently viable osteocytes with human stained nuclei present in lacunae with corresponding PDPN at 12 weeks (Figures 4D, E, and S7, Supporting Information). A distinct variance in the regenerated bone tissue morphology was evident between the tested groups at the defect region. Consistent with SEM findings, there was no substantial tissue formation observed in the empty defect, and only fibrous connective tissue formation in the non-mineralized construct-treated groups. H&E staining showed the presence of inflammatory cells, particularly, a mixture of plasma cells, lymphocytes, and bone-resorbing multi-nucleated macrophage-like cells, located primarily around the implanted cell-laden mineralized constructs (Figure S7B, Supporting Information). Both H&E and trichrome staining revealed the initiation of well-integrated bone apposition by BMP-2 and cell-laden mineralized samples at the periphery of the defect site (demarcated by arrowheads, Figure 4A–C, Figure S7, Supporting Information). An apparent transformation of the injected cell-laden mineralized constructs into mature bone tissue was seen across the defect, wherein viable osteocytes and distinct osteoblasts rimming were apparent (Figure S7, Supporting Information), with evidence of resting lines around the implanted material (Figure 4B). Trichrome staining also demonstrated evidence of remodeling of cell-laden mineralized constructs at the core of the defect into woven bone (disorganized collagen fibers) and lamellar bone with diploe, osteoblasts, and osteocytes (Figure 4C). The cell-laden mineralized group demonstrated bone remodeling and maturation both circumferentially and throughout the body of the implanted constructs, as evidenced by mature osteocytes in lacunae (Figure S7, Supporting Information). The cell-laden mineralized group was also found to have more vascular invasion and cell penetration around the implanted constructs (Figure 3E and I). Moreover, the average thickness of regenerated tissue at the defect was in the order of cell-laden mineralized > BMP-2 > non-mineralized > empty (Figure 4F).

Figure 4. Functional bone regeneration is regulated by viable cells within mineralized cell-laden collagen microgels.

Tile scan images of A) H&E stained histological sections of bone tissue samples for different treatment groups at 12-weeks post-transplantation. Arrowheads in the left and right sides of the images demarcate the native bone versus regenerated bone. Magnified images of dashed box regions show a detailed description of new bone formation within the implanted constructs at the peripheral region of the defect. B) A magnified tile scan image reveals the transformation of mineralized constructs into the new bone at the core region of the defect. H&E staining shows bone as a compact structure in a dark pink color, the connective tissue in light pink, and well-defined structures of implanted constructs, bone remodeling, bone cells, osteoid matrix, and blood cells. $ indicates non-mineralized constructs, # indicates mineralized constructs, and * indicates the newly formed bone from mineralized constructs. C) Masson’s trichrome stained Masson’s trichrome shows general woven bone tissue in dark blue, and collagen matrices in light blue. Immunofluorescence staining of D) HuNu and E) podoplanin at the core region of the defect. F) The average tissue thickness (n = 4) of regenerated bone tissue in the defect site. A positively stained area was quantified for G) HuNu (n = 4) and H) PDPN (n = 4) via ImageJ. An ordinary one-way ANOVA with Tukey’s multiple comparison test was used. Error bars are mean with s.e.m., and the significant effect of different groups on bone regeneration is denoted as * (p < 0.05), ** (p < 0.01), *** (p < 0.001), and **** (p < 0.0001), respectively.

Immunofluorescence staining of human-specific nuclear antigen (HuNu) was performed to confirm the presence of the originally implanted hMSCs embedded within the cell-laden mineralized constructs after 12 weeks (Figure 4D). Human cells (HuNu+) were observed only in the cell-laden non-mineralized and cell-laden mineralized groups, with the cell-laden mineralized constructs containing 1.6-fold more HuNu positive area (Figure 4G). Additionally, the morphology of cells in cell-laden mineralized constructs was visibly different from non-mineralized samples (Figure S8C, 6 week post-transplantation, Supporting Information). This result indicates that the hMSCs remained surprisingly viable and remained in the site of regeneration six and twelve weeks after implantation via our encapsulation method, with an increased cell retention rate with cell-laden mineralized constructs (Figure 4D, G, and S8C, Supporting Information). Immunofluorescence staining of podoplanin (PDPN), an early-stage osteocyte marker, was also used for the detection of osteocytes. PDPN expression was observed in BMP-2 and cell-laden mineralized groups (Figure 4E). Notably, PDPN staining was homogeneously distributed throughout the cell-laden mineralized and BMP-2 groups comparable to that of the surrounding native bone. Images demonstrated a significantly increased expression of PDPN for BMP-2 and cell-laden mineralized groups in comparison to that of empty and non-mineralized groups (Figure 4H) at 12 weeks and similarly 6 weeks post-operatively (Figure S8D, Supporting Information).

The bone deposition patterns of the cell-laden mineralized and acellular mineralized constructs were starkly different. In H&E and Masson’s trichrome staining, acellular mineralized constructs only demonstrated bone deposition along the peripheral aspects of the constructs (Figure S9, Supporting Information). In immunofluorescence staining, there was no presence of HuNu+ cells, and PDPN was only noted along the periphery of each construct. This finding indicates that the osteocytes present were host-derived, and that remodeling of the acellular mineralized construct occurred from the periphery inwards. When comparing the cell-laden mineralized and acellular mineralized groups, the incorporation of hMSCs within the mineralized constructs led to more robust regeneration, as both graft-derived and host-derived cells appear to contribute towards the formation of mature bone tissue.

3. Discussion

Here we present an engineered bone-like injectable cell-laden biomaterial able to promote new bone formation, without the disadvantages that are inherent to recombinant growth factor supplementation or autologous bone harvesting. Furthermore, we demonstrated that this engineered injectable constructs replicate the nano-scale characteristics of mineralized bone,^[10] which preserved the viability and engraftment of hMSCs in vivo after at least 12 weeks post-implantation. We also demonstrate that the nano-scale calcification of hMSCs in collagen leads to the upregulation of osteogenic genes via the action of Rho/ROCK mechanotransduction mediated effects and adenosine signaling, possibly in an interdependent or compensatory manner, which may shed new light on the process of native intramembranous ossification of stem cells. Overall, this report demonstrates, for the first time, the unique ability of a cell-laden, bone-like, calcified biomaterial to be minimally injected and regenerate bone tissue, potentially avoiding the disadvantageous effects of growth factor supplementation.

In our mineralized constructs, the embedded cells experience a rapid stiffness increase at the level of individual cell-binding sites, due to the nano-scale intrafibrillar calcification of individual fibrils (Figure S3J, Supporting Information).^[10] This nano-scale stiffening of the matrix at the adhesion sites requires the assembly of focal adhesions and force-dependent cytoskeletal components such as vinculin, which are notably absent in soft, pliable substrates,^[18] in order to stabilize cell response to the additional force/tension experienced by the cells. Tensional homeostasis is accomplished by a concomitant increase in cytoskeletal tension typically associated with the formation of stress fibers by RhoA-mediated ROCK activation.^{[13, 19]} The selective inhibition of ROCK (via Y27632) decreased RUNX2 expression in the mineralized constructs, indicating the role of mechanotransduction on osteogenic differentiation without any osteogenic inducing factors (Figure 2C, D, G, and H). Concurrently, the ROCK inhibitor (Y27632) had no significant effects on the osteogenic differentiation of non-mineralized constructs cultured in OIM, indicating the absence of mechanotransduction (stiffness) driven osteogenesis as found in our mineralized constructs and bone. Together, these results suggest that our nano-scale mineralization protocols closely approximate the native mechanisms of cell differentiation found in native osteogenesis, whereas conventional supplementation with the osteoinductive medium is poorly affected by these known mechanoregulatory mechanisms.^[8] Interestingly, when we utilized an A2b receptor inhibitor PSB603 (Figure 2I),^[15b] the RUNX2 expression of hMSCs in the mineralized constructs also decreased significantly, indicating the role of ion channels on osteogenic differentiation, (Figure 2E and G) irrespective of the changes in substrate mechanics. Our results agree with the previous findings, where phosphate ions are transported into the cytoplasm via SCL20a1 transporter to produce ATP intracellularly, promoting osteogenic differentiation.^[20] Moreover, given the statistically comparable reduction in differentiation markers via inhibition of either the mechanotransduction or ionic-dependent pathways suggests that these mechanisms may be interdependent. In other words, blocking of the phosphate-dependent pathway with preservation of the mechanotransduction pathway was insufficient to maintain the same levels of RUNX2 expression observed without inhibitory treatments. At the same time, blocking of mechanotransduction via the ROCK inhibitor with preservation of adenosine signaling had a statistically comparable effect. Further studies are required to elucidate the participation of these mechanism in mineralization-induced osteogenic differentiation.

Histological analysis (Figure 4) suggested that both the implanted hMSCs and native host cells contributed to bone remodeling. The identification of viable HuNu positive cells within the mineralized constructs even after 12 weeks further demonstrated that regeneration was facilitated substantially by the human cells embedded within the constructs. PDPN expression data also indicated that hMSCs within the mineralized constructs initiated osteocyte differentiation, in addition to the cells native to the recipient site. Traditionally, implanted cells have had a transient effect to facilitate regeneration, but rarely have been reported to engraft within the host bone to mature phenotypically and express markers that are suggestive of long-term survival and function. Interestingly, high expression of MMP8 and MMP13, as well as BMP4 and IGF, all of which have been linked with bone matrix remodeling and osteoclast activation,^[21] suggests that cells within the mineralized samples may actively regulate scaffold remodeling in a paracrine fashion, similar to the known participation of osteocytes in native bone.^[22] Accordingly, remodeling occurred throughout the entire defect region, unlike acellular mineralized constructs, where remodeling occurred only around the periphery of each construct by the host cells (Figure S9, Supporting Information). The cell-laden mineralized constructs had significantly more blood vessels formation than the other groups, indicating that they may also have an innate angiogenic potential. Overall, our in vivo studies indicate that the mineralized constructs not only mimicked the structure of native bone with nano-scale precision, which we had previously demonstrated, but also that this precision engineering of the matrix where cells are embedded can mimic the pattern of bone regeneration that is observed in autologous bone grafts. High magnification H&E staining of our mineralized constructs revealed mature lamellar bone with osteocytes in lacunae analogous to the remodeling pattern previously reported in autologous bone fragments (Figure S7, Supporting Information),^{[16a, 17]} where remodeling of autologous bone occurred primarily in the periphery of the fragments, with focal points of mineralization within the center. This pattern indicated that remodeling occurred both from cells present at the recipient site, and through cells naturally present in the donor material.^[17] Interestingly, the increase in BV measured via microCT, from 0.07 mm³ at week 0 to 25.4 mm³ in week 12, also indicates a poorly understood mechanism where newly formed bone is accumulated via complete remodelling (degradation) of the engineered matrix on the edges of the implanted microgels, whereas the cell-laden mineralized matrix may continue to mineralize in vivo in a process akin to the maturation of woven bone, which is visible in the longitudinal analyses of our microCT data (Figure S6). These possible mechanisms will require further analyzes for complete clarification. Taken together, the long-term engraftment of implanted hMSCs, the expression of mature bone cell markers, combined with the angiogenic potential and remodeling potential of these injectable microgels, may point to a novel class of regenerative bone biomaterials that can rival the regenerative potential of growth-factor dependent bone grafts.

4. Conclusion

Findings from our in vivo study demonstrated that our engineered injectable constructs are inherently osteogenic and capable of regenerating bone, despite the absence of any growth factors, as evidenced by substantial bone volume and histological presence of active bone modeling and remodeling processes, despite the absence of stimulating growth factors or exogenous supplements. Moreover, our results demonstrate that the embedded hMSCs are viable and contribute to enhanced osteogenesis during healing 12 weeks after implantation. Therefore, these biomimetic engineered constructs may have future potential for robust bone formation as an alternative to current treatment strategies for bone, dental, and craniofacial tissue repair.

5. Materials and Methods

Experimental design:

The goal of this study was to investigate the regenerative potential of nano-scale mineralized cell-laden collagen constructs in comparison to therapeutic growth factor (BMP-2) using well established critical-sized calvarial defect model.^[16] The osteogenic potential of our system was evaluated both in vitro using an osteogenesis gene array kit and in vivo using noninvasive μCT imaging and histology. All animals and treatment groups were assigned at random, and investigators were blinded for all in vitro, in vivo, and ex vivo analyses.

Cell culture:

Human mesenchymal stem/stromal cells (hMSCs, Cell Applications, Inc., Cat # 492k-05a, Lot # 2758) were used from passages 2 to 5 for all experiments. Cells were cultured in alpha modified eagle medium (α-MEM, Gibco) supplemented with 10% fetal bovine serum (FBS, HyClone), 1% L-Glutamine (200 mM), and 1% antibiotic solution (Corning) at standard culture condition (37 °C and 5% CO₂). Cells were sub-cultured using HyClone^™ Trypsin Protease (Thermo Fisher Scientific) when reaches more than 80% confluence.

Non-mineralized constructs:

Non-mineralized collagen constructs were prepared as previously described.^{[16, 23]} Acid solubilized Type 1 collagen from rat tail (3 mg/mL BD Biosciences) was reconstituted in phosphate-buffered saline (PBS, 10X, Gibco) and cell culture growth medium for a final concentration of 1.5 mg/mL. The collagen solution’s pH was neutralized using 1N sodium hydroxide (NaOH, Sigma-Aldrich) solution, sterilized using a 0.22 μm pore filter (Fisher Scientific). Then, hMSCs were mixed with the abovementioned collagen solution to make a final cell density of 5 X 10⁵ cells/mL. The resulting collagen solution containing hMSCs was used for the non-mineralized cell-laden collagen constructs preparation. The entire procedure was carried out in an ice bath to prevent the gelation of collagen. Then 100 μL of the prepared cell-mixed collagen solution was cast onto six-well plates and allowed for fibrillogenesis in a humidified 5% CO₂ incubator at 37°C for 15 min.

Mineralized constructs:

Cell-laden collagen constructs were prepared as described above. A modified mineralization medium was used to optimize the nano-scale mineralization of cell-laden collagen constructs (Figure 1A).^[10] Briefly, Dulbecco’s Modified Eagle Media (DMEM, Thermo-Fisher Scientific) was supplemented with 10% FBS, 1% antibiotic solution, and 25 mM 2-[4-(2-hydroxyethyl)piperazin-1-yl]ethanesulfonic acid (HEPES, Fisher Scientific). The pH of the resulting media was neutralized using 1N NaOH. HEPES was used for the stable maintenance of pH. The resulting media was divided equally to prepare 9 mM calcium chloride (CaCl₂·2H₂O, J.T. Baker) and 4.2 mM potassium phosphate (K₂HPO₄, J.T. Baker). Osteopontin (OPN-10, Lacprodan^®; Arla Foods Ingredients Group P/S, Denmark) was dissolved in a culture medium for a 100 μg/mL concentration and filtered using a 0.22 μm pore filter. Osteopontin instructed the nano-scale mineralization of collagen fibrils. To achieve adequate mineralization, osteopontin solution was mixed thoroughly with the media containing CaCl₂, followed by mixing thoroughly with the media containing K₂HPO₄. The mineralization media was prepared freshly before the use. The well plate containing cell-laden collagen constructs was placed in a rotary shaker and incubated at standard culture conditions, and the mineralization medium was replenished every 24 hours for a total of three days.

Preparation of micro-sized constructs:

Injectable micro-sized non-mineralized and mineralized constructs were prepared using a micro-sectioning technique (The McIlwain^™ Tissue Chopper). Briefly, the equipment was sterilized by ultraviolet light in the biosafety cabinet. A double-edged blade (Wilkinson Sword) attached to the chopper arm was positioned to ensure a horizontal approach on the plastic disc (ethanol sterilized and dried) placed on the platform. The monolithic construct was placed on the disc and micro-sectioned at a stroke of 90/min to prepare micro-sized constructs. Vernier calipers were adjusted on the tissue chopper to get the micro-sized constructs sizes ranging from 0.1 to 1 mm². Unlike soft hydrogels, mineralized collagen gel was easy to handle (Figure S2 A and B, Supporting Information) due to the increased mechanical stiffness conferred by the calcification. We used 0.9 mm² sized mineralized collagen microgels for the in vitro and in vivo experiments. The prepared micro-sized constructs were placed in the 6 well plates and incubated in standard culture conditions until further use. A live and dead cell assay (Molecular Probes) was performed to determine the cell viability of micro-sized constructs.

Analyses of biomimetic nano-scale mineralization:

A longitudinal analysis was performed to determine the complete calcification of the mineralized constructs. Briefly, the constructs (n = 3) were collected at 0 hr, 12 hr, 24 hr, 36 hr, 48 hr, 60 hr, and 72 hr, respectively, for Alizarin red S staining, von Kossa staining, and Fourier-transform infrared spectroscopy (FTIR) analysis. Samples (n = 3) collected after 72 hours were used to analyze mechanical properties and morphological characteristics.

Alizarin Red S and von Kossa staining were performed to evaluate calcium and phosphate deposition, respectively. The collected constructs were washed with 1X PBS then fixed with 10% neutral buffered formalin (Thermo Scientific) for 30 min at room temperature. For Alizarin Red S staining, the samples were washed with deionized (DI) water twice and treated with 2% Alizarin Red S solution (pH 4.1–4.3, Electron Microscopy Sciences) for 15 min at room temperature. For von Kossa staining, the fixed samples were washed with DI water twice and treated with 5% silver nitrate (AgNO₃, J. T. Baker) under UV light for 1 h. Afterward, the samples were washed with DI water three times and treated with 5% sodium thiosulfate (Na₂S₂O₃, Sigma-Aldrich) for 1 min. The samples were then washed with DI water and treated with 5% sodium thiosulfate for 5 min to remove un-reacted silver. Stained samples were subsequently washed with DI water and imaged at 4x magnification in an inverted fluorescence microscope (FL Auto, EVOS).

FTIR was performed to determine the functional groups and structural characteristics of the samples. Longitudinally collected samples were washed in DI water thrice and lyophilized (Labconco Freeze Dry System). The samples were stored in a vacuum chamber to ensure complete drying. FTIR spectra were collected in transmission mode (Nicolet 6700, Thermo Scientific) using 32 scans in the range of 4000–400 cm⁻¹ at a resolution of 4 cm⁻¹. The mineral to matrix ratio was calculated from the area under the curve of ν₃PO₄ (1030 cm⁻¹) over amide (1660 cm⁻¹) peaks after baseline correction and normalization. All the height measurements were performed using Omnic Spectra software after baseline correction and normalization of the spectra to the amide I band (1585–1720 cm⁻¹).

The compressive modulus of the non-mineralized and mineralized monolithic constructs was measured by unconstrained compression tests using a mechanical testing machine (MTS Criterion Model 42, MTS Systems Corporation, MN, USA) equipped with a 100 N load cell. The loading rate was kept at 0.01 mm/sec on samples using the construct base’s cross-sectional area.

SEM, TEM, EDAX, and SAED were performed to evaluate the morphology, elemental analysis, and crystallinity of the samples, respectively. For SEM, the samples were dehydrated through a graded series of ethanol (25% to 100%), critical point dried, sputter-coated with platinum, and imaged via SEM (QuantaTM, FEI, and FEI Helios Nanolab^™ 660 DualBeam^™) to determine the deposition of minerals on the collagen fibrils. Elemental analysis was performed with an EDAX detector (INCA, Oxford Instruments). For TEM, the mineralized collagen constructs (72 hr) were minced and immersed in ice-cold 0.1M ammonium bicarbonate solution. The minced construct was homogenized, and the homogenate was pipetted onto freshly glow-discharged 600-mesh carbon-coated TEM grids. The samples were imaged through TEM (Tecnai G20, FEI) operated at 120 kV. SAED analysis was performed using lyophilized samples in a TECNAI F20 TEM with an Oxford SDD EDS detector and Gatan GIF 2001 system operated at 200 kV. Non-mineralized collagen construct was used as a control group.

Cell behavior:

The cells’ viability in collagen construct after mineralization was evaluated using a live and dead assay kit (Molecular Probes). The cell-laden collagen constructs (n = 3) were cultured in a growth medium and mineralization medium for three days. The live and dead cell assay was performed following the manufacturer’s instructions, and the samples were imaged in a laser-scanning confocal microscope (Zeiss LSM 880). For the proliferation assay, the number of cells before and after mineralization was measured by incubating the samples (n =3) with a fresh cell culture medium containing 10% v/v AlamarBlue (Bio-Rad) for 5 hours. The resulting fluorescent resazurin products in the culture medium were measured in 96-well plates using GloMax^® Explorer (Promega) at excitation and an emission wavelength of 580–640 nm and 520 nm, respectively. A standard curve of known cell numbers was used to determine the samples’ cell numbers by correlating the fluorescent readings. Non-mineralized collagen construct was used as a control group. Immunofluorescence staining of the focal adhesion molecule (vinculin) co-stained with the cytoskeleton (F-actin) and nucleus was performed to determine the cell spreading, aspect ratio, and focal adhesion assembly pattern between non-mineralized and mineralized collagen samples. After three days of culture, the constructs were fixed with 4% paraformaldehyde for 30 min at RT, permeabilized with 0.1% Triton X-100 for 10 mins, blocked with 1.5% bovine serum albumin (BSA, Sigma-Aldrich) for 1 hr at RT, incubated with monoclonal anti-vinculin antibody produced in mouse (V9131, Sigma-Aldrich, 1:250) at 4 °C overnight, incubated with Alexa fluor 647 goat anti-mouse (Invitrogen, A21235, 1:500) for 2 hours at 4 °C, co-stained with NucBlue^™ Fixed Cell ReadyProbes (Invitrogen, Thermo Fisher Scientific) for nucleus and ActinRed^™ 555 (ReadyProbes^™, Invitrogen, Thermo Fisher Scientific) for F-actin. The samples were rinsed in 1x PBS thrice between each step listed above. The samples were then imaged in a laser-scanning confocal microscope (Zeiss LSM 880). The spreading area and aspect ratio of cells encapsulated in the constructs were determined using ImageJ (Fiji).

Osteogenic differentiation:

Three groups (n = 7) were tested and cultured for a total of seven days. The non-mineralized constructs were cultured in 1) growth medium (negative control), 2) osteogenic induction medium (OIM, positive control), and 3) mineralization medium for three days, then subsequently cultured in growth media for an additional four days (Figure S5A, Supporting Information). For OIM, the growth medium was supplemented with 100 nM dexamethasone (Sigma–Aldrich), 10 mM β-glycerophosphate (Sigma–Aldrich), and 50 μM ascorbic acid (Sigma–Aldrich).

Osteogenesis gene array (Reverse transcription-polymerase chain reaction):

Non-mineralized and mineralized constructs were cultured under three different medium conditions for 14 days and were evaluated for gene expression. Briefly, RNA was extracted from collagen constructs (n = 6) using TRIzol according to the manufacturer’s instructions. 1 mg of RNA was reverse-transcribed to complementary DNA (cDNA) using an iScript cDNA synthesis kit (Bio-Rad, catalog no. 170–8891) for each collagen construct. Real-time PCR reactions were conducted on ABI Prism 7700 Real-time PCR Cycler (Applied Biosystems). Osteogenic differentiation of hMSCs was examined using a human Osteogenesis PCR array (SABiosciences, catalog no. PAHS-026). A total of 68 genes were analyzed for PCR array, and their relative expressions were normalized to collagen constructs cultured in growth media, and the data were presented as a heat map. The colors of the heat map indicate expression from low (blue) to intermediate (white) to high (red).

Mechanistic study:

For pharmacological inhibition of adenosine receptors, the media was treated with A2b receptor inhibitor, 8-[4-[4-(4-Chlorophenzyl)piperazide-1-sulfonyl)phenyl]]-1-propylxanthine (PSB603, Tocris Bioscience) at a concentration of 100 nM and the media was changed every 2 days, except for mineralization group where media was changed every day. For pharmacological inhibition of ROCK, the media was treated with 4-[(1R)-1-aminoethyl]-N-4-pyridinyl-trans-cyclohexanecarboxamide, dihydrochloride at a concentration of 30 μM (Y-27632, Sigma-Aldrich), and the media was changed every day. The samples were then prepared for immunofluorescence staining of RUNX2 and YAP as described above using polyclonal anti-RUNX2 antibody produced in rabbit (NBP1–77461, Novus Biologicals), monoclonal anti-YAP1 produced in mouse (WH0010413M1, Sigma-Aldrich) respectively with Alexafluor 488 goat anti-rabbit (Invitrogen, A11008, 1:500), Alexa fluor 647 goat anti-mouse (Invitrogen, A21235, 1:500) and NucBlue^™ Fixed Cell ReadyProbes (Invitrogen, Thermo Fisher Scientific). Samples were then imaged in a laser-scanning confocal microscope (Zeiss LSM 880), and the intensity of RUNX2 and YAP expression was quantified for all the cells in a sample using ImageJ (Fiji). The average intensity associated with the nuclear RUNX2 and YAP was normalized against the cell number in each sample (n = 7).

Surgical procedures:

All animal procedures were performed (Figure S6A–D, Supporting Information) in strict accordance with the protocols approved by the OHSU Institutional Animal Care and Use Committee (IACUC protocol #IP00000570 - Implantation of pre-vascularized and mineralized tissue-engineered scaffolds in rat calvarial defects). A total of 21 Wistar rats (9 weeks old; Charles River Laboratories) were acclimated for two weeks before any procedures were performed. Rats were double-housed in ventilated caging with pelleted cellulose bedding, brown crinkled paper, red polycarbonate tunnel, and nylon gnawing bones for enrichment. The rats were housed on a 12-h light/dark cycle and allowed ad libitum access to food (PicoLab 5LOD, LabDiet, St. Louis, MO) and reverse osmosis, chlorinated, autoclaved water in bottles. Four experimental groups were tested: empty defects (negative control, n = 4), BMP-2 loaded collagen constructs (positive control; 2.5 μg/defect, n = 5), non-mineralized constructs (n = 6), and mineralized constructs (n = 6). BMP-2 loaded collagen constructs were prepared as previously.^[24] Pelaez et al., tested the dose-dependent effect of BMP-2 on bone regeneration using a preclinical calvarial bone defect model in rodents.^[5c] The study demonstrated that BMP-2 between 1.25 μg - 2.5 μg achieved 90% bone regeneration in the calvarial defect model. Moreover, the study claimed that no additional further enhancement to bone formation/maturation was observed for BMP-2 above the 2.5 μg dose. In addition, severe side effects, such as inflammation and heterotopic bone formation, have been associated with high-dose BMP-2 delivery,^[6a] indicating an appropriate dose is crucial to achieving effective bone regeneration without side effects.^[5a] Therefore, to achieve effective bone regeneration while minimizing the side effects and to omit additional variation in the test group, which is the clinical objective in all cases, we opted to use 2.5 μg BMP-2 in this study as a positive control. A critical-sized calvarial defect (~ 8 mm Ø) was created as previously described.^[16] Briefly, rats were put under general anesthesia through isofluorane (Henry Schein Animal Health, Dublin, OH, USA) inhalation using a nasal hood. A sagittal incision was made from the midpoint between the canthi posteriorly to the occipital bone. The dermis and periosteum were reflected by blunt dissection to expose the calvarium. An 8 mm diameter trephine bur was used on a contra-angle low-speed dental handpiece to create the defect under continuous saline irrigation; the defect was centered over the intersection of the coronal and sagittal sutures. The island of bone was removed without disrupting the underlying dura. 250 μL of the non-mineralized or mineralized constructs with 225 μl volume of carrier fluid were injected into the defect using a syringe; test groups were allocated based on a randomization protocol. For injection, the microgels were loaded into a syringe in cell media as the solution of suspension without the use of any other delivery carrier. The dermis was closed over the defect utilizing surgical staples. Each rat received a subcutaneous injection of Meloxicam (OstilLox^™) post-operatively for analgesia. The animals were monitored daily post-operatively for seven days for signs of complications at the surgical site or abnormal behavior. Surgical staples were removed two weeks post-operatively and animals were euthanized through carbon dioxide asphyxiation 12 weeks post-implantation.

Analyses of bone regeneration:

Longitudinal bone formation was qualitatively and quantitatively evaluated using three-dimensional μCT imaging at 0 week (Viva-CT 40, Scanco Medical), 4 weeks (Inveon, Siemens), 8 weeks (Inveon, Siemens), and 12 weeks (Viva-CT 40, Scanco Medical) post-treatment. MicroCT was performed with X-ray Settings E = 55 kVp, I = 145 μA, power 8 W, and a voxel size of 38.9 μm, and the bone volume (BV) and bone mineral density (BMD) were quantified by Scanco software, following our established protocol.^[25] Obtained microCT data were contoured and evaluated over 160 slices from a central region within the bone defect to normalize total evaluative volume between test groups without including intact native bone. The BV data at 12 weeks for mineralized cell-laden microgels treated groups were subtracted with BV data measured at 0 week. The percentage of the bone bridging area was analyzed using ImageJ (FIJI), as established previously.^[16a] After euthanasia, the calvarium was dissected, processed, fixed in 10% neutral buffered formalin (Thermo Scientific), and decalcified at room temperature, as established previously.^[16a] The tissue samples were dehydrated in a graded series of ethanol (from 25% to 100%), embedded in paraffin, serially sectioned into 7 μm thick sagittal sections, and stored for subsequent analysis. For SEM imaging, sagittal in vivo sections were dehydrated through a graded series of ethanol (25% to 100%), dried completely, sputter-coated with platinum, and imaged via SEM (QuantaTM, FEI) to determine the newly formed bone, remodeling of mineralized and non-mineralized constructs, and vascular network formation within the defect region.

The sectioned samples were subjected to hematoxylin and eosin (H&E) and Trichrome staining at the Histopathology core (OHSU) following standard in-house protocols. For immunofluorescence staining, the samples were washed three times with PBS, antigen retrieved, blocked with 3% BSA for 2 hours, and incubated individually with the following respective primary antibodies overnight at 4 °C: alpha-smooth muscle actin (α-SMA, mouse monoclonal anti-αSMA, Invitrogen, MA5–11547, 1:400 dilution), nuclear antigen-antibody specific to human (HuNU, mouse monoclonal anti-HuNU, Novus, NBP2–34342, 1:200 dilution), and podoplanin (PDPN, mouse monoclonal anti-PDPN, Origene, DM3500P, 1:100 dilution). The samples were then washed with PBS and incubated with the corresponding secondary antibody overnight at 4°C: Alexafluor 488 goat anti-mouse for HuNu and PDPN (Invitrogen, A11001, 1:500), or Alexafluor 647 goat anti-mouse for α-SMA (Invitrogen, A21235, 1:500). The nuclei were stained with 4′,6-diamidino-2-phenylindole special formulation (DAPI, NucBlueTM Fixed Cell Stain ReadyProbes, Invitrogen, Thermo Fisher Scientific). The stained tissue sections were imaged using an inverted fluorescence microscope (FL Auto, Evos) or a laser-scanning confocal microscope (Zeiss LSM 880). The positive immunofluorescence staining areas were quantified from three randomly selected images per group using ImageJ (FIJI).

Statistical analysis:

A student’s t-test was used to compare the compressive modulus, cell spreading area, actin, and vinculin expression of non-mineralized and mineralized groups. An ordinary one-way ANOVA with Tukey’s multiple comparison test was used to compare longitudinal mineral:matrix ratio (n = 3), gene expression (n = 6), RUNX2 expression (n = 7), YAP expression (n = 7), BV (n = 4), bone bridging area (n = 4), BMD (n = 4), HuNu+ staining area (n = 4), PDPN+ staining area (n = 4), and αSMA+ staining area (n = 4). An ordinary two-way ANOVA with Tukey’s multiple comparison test was used to compare cell viability. All statistical analyses were performed in GraphPad Prism 9 with α=0.05. Data are presented as mean ± standard error of the mean.