Cryogel scaffolds (n = 4/group/timepoint) were fabricated based on a previously established protocol (Kathuria et al., 2009). Briefly, 80 mg of low-viscosity chitosan (MP Biomedicals) was dissolved in 8 ml of 1% acetic acid solution (Fisher Scientific). Following dissolution, 320 mg of gelatin (from cold water fish skin; Sigma) was added to the solution and mixed for 1 h. In a separate scintillation vial, 1% glutaraldehyde solution was prepared using 2 ml of 1% acetic acid solution for cross-linking. All the solutions were cooled at 4°C for 1 h. The solutions were mixed by decanting and then immediately poured into precooled 12-well plates (Sigma), such that a thickness of 3–4 mm was achieved, and frozen at −20°C for 18 h. At this time, the plates were thawed at room temperature (RT) water and trimmed into rectangular sheets (6 mm × 12 mm × 1.5 mm). The scaffolds were sterilized in 1% peracetic acid (PAA; Sigma) for 90 min and then rinsed with sterile 1X phosphate-buffered saline (PBS) three times for 10 min (Yoganarasimha et al., 2014; Costa et al., 2015). Following sterilization, all the scaffolds were placed at −80°C for 1 h and subsequently lyophilized (Sentry 2.0 VirTis BenchTop Pro Freeze Dryer; SP Scientific) for 24 h to ensure complete drying. The resulting cryogel scaffolds were macroporous (pore diameters of 30–100 μm) and sponge-like in structure, allowing them to undergo compression without the crack propagation (Brunger et al., 2014; Gómez-Barrena et al., 2015). Additionally, the cryogels were fabricated from the natural materials that render them compatible with cells and supportive of angiogenesis (Glass et al., 2014).

The doxycycline-inducible lentiviral vector was generated by cloning BMP2 into the modified TMPrtTA vector (provided by Danos Lab) as previously described (Rowland et al., 2018). To produce the lentivirus, HEK293T cells were co-transfected with either the dox-BMP2 vector or a constitutive GFP lentiviral vector with a second-generation packaging plasmid psPAX2 (Addgene, No. 12260) and an envelope plasmid pMD2.G (Addgene, No. 12259) by calcium phosphate precipitation (Salmon and Trono, 2007). The lentivirus was stored at –80°C until further use. The functional titer of each virus was determined via qRT-PCR to ascertain the number of lentiviral DNA copies integrated into the genome of the transduced HeLa cells (Salmon and Trono, 2007). On the day of transduction, the lentivirus was thawed and concentrated ∼100-fold using Amicon Ultra 100 kDa MWCO filters (Millipore, UFC9100).

To immobilize the lentivirus on the scaffolds, cryogels (n = 4/group/timepoint) were incubated overnight with 0.002% poly-L-lysine (Sigma), washed with PBS, and then incubated with enough concentrated lentivirus to transduce 500,000 or 1 million cells for 1.5 h at 37°C. The scaffolds were rinsed again with PBS, moved to a new low-attachment 24-well plate (Corning), and seeded with 500,000 cells for in vitro experiments and 1 million cells for implanted scaffolds. The expansion medium was added to each sample for 1 h after the cell seeding. All mediums were changed 24 h later. The scaffolds were cultured in the expansion medium for six days to facilitate cell infiltration. A subset of scaffolds were then fixed in a neutral-buffered formalin (NBF) for 24 h, placed in a 30% sucrose solution for an additional 24 h, and embedded in an optimal cutting temperature (OCT) compound. Once frozen, the cryogels were cryosectioned at 20 µm and stained with 4′,6-diamidino-2-phenylindole (DAPI) to visualize the cellular infiltration. All other scaffolds remained in culture.

After 6 days of culture in the expansion medium, the medium was changed to osteogenic medium containing DMEM-HG (Gibco), 10% lot selected FBS, 1% penicillin–streptomycin (Gibco), 50 μg/ml L-ascorbic acid 2-phosphate (Sigma), 10 nM dexamethasone (Sigma), and 10 mM β-glycerophosphate (Sigma). Doxycycline (dox) groups were supplemented with either 0.1 or 1 μg/ml dox (Sigma). Half medium changes were performed every other day for 28 days, and medium samples were collected on days 0, 2, 6, 10, 16, 22, and 28 of the osteogenic culture. The concentration of BMP-2 in the medium (n = 4/group/timepoint) was measured by ELISA (R&D).

Following 28 days of osteogenic culture, the scaffolds (n = 4/group/timepoint) were scanned using the Bruker SkyScan 1176 micro-CT at 40 V, 600 µA. Micro-CT datasets were reconstructed with NRecon software using a dynamic range of 0.1, ring artifact correction of 11, and beam hardening correction of 20%. Mineralization was quantified via calibration against hydroxyapatite phantoms using CT-Analyzer software. Acellular scaffolds soaked in the medium for 28 days served as a control.

A total of 62 male and female mice at 12 weeks of age were used in one of the four experimental groups: sham, scaffold only, GFP cells + scaffold, and BMP-2 cells + scaffold. Experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at Washington University in St. Louis in accordance with the Animal Welfare Act and PHS Policy on Humane Care and Use of Laboratory Animals. Atrophic non-unions were established in transgenic 3.6Col1A1-tk (Col1-tk) mice (rederived from the frozen embryos provided by Drs. Robert Jilka and Charles O’Brien) by targeting the replicating osteoblast progenitors (Jilka et al., 2009). As these mice express the herpes simplex virus thymidine kinase gene (HSV-tk, or “tk”), driven by the 3.6-kb rat Col1A1 promoter which is active in the osteoblast lineage cells; the addition of the nucleoside analog ganciclovir (GCV) targets replicating osteoblast progenitors (Figure 1A). This process occurs through the conversion of GCV to a toxic version of the nucleotide which is incorporated into the DNA, resulting in strands breaking and cell apoptosis (Jilka et al., 2009). To generate Col1-tk mice, male C57BL6/J (The Jackson Laboratory, #000664) mice were bred with female mice, heterozygous for the tk transgene (tk-positive). This resulted in the heterozygous tk-positive (Col1-tk) and tk-negative (wild-type, WT) mice. Only one allelic copy of the tk transgene is necessary to achieve the tk-positive and the male Col1-tk mice are sterile. All genotyping procedures were completed by Transnetyx (Cordova, TN) from toe biopsies taken from the mice for the real-time PCR (probe: puro). Additional breeding was used in the littermates (tk-positive females and tk-negative males). Prior to and during the experimental procedures, all the mice were group-housed (up to five mice/cage) under standard 12-h light/dark cycles with full access to food and water. Note that breeders were given high-fat chow, but following weaning, the experimental mice were given normal chow. All the mice were dosed with ganciclovir (GCV, 8 mg/kg i.p., McKesson) once daily starting at the fracture (12 weeks of age) for 2 weeks. The mice were euthanized by CO₂ asphyxiation at designated endpoints from 6 to 12 weeks after the fracture.

We recently demonstrated a model of atrophic non-union in Col1-tk mice dosed with GCV (Hixon et al., 2021). Briefly, a previously established protocol was employed to create a full fracture in the right femur of 12-week-old mice (Gardner et al., 2011). All mice were given buprenorphine SR (1 mg/kg, s.c.) for analgesia 1 h before surgery. The fracture was created by applying a transverse force across the right femoral mid-diaphysis using a custom 3-pt bending setup (DynaMight 8841; Instron). The femur was then stabilized using a 24-gauge metal pin (MicroGroup), and the incision was closed using 3‐0 nylon sutures (Ethicon). We confirmed proper fixation by radiography (Faxitron Ultrafocus 100; Faxitron Bioptics) immediately following the fracture surgery. Note that in the WT mice, this model closely mirrors the classic fracture injury and healing response (Einhorn, 1998; Bonnarens and Einhorn, 1984), displaying both endochondral and intramembranous healing within the fracture gap and around the periphery, respectively (Liu et al., 2017; McKenzie et al., 2019). To establish the atrophic non-union, Col1-tk mice were dosed with GCV at the time of bone injury for 2 weeks, at which time the drug was withdrawn. All the animals were then monitored by weekly radiography (Faxitron) for an additional 2 weeks to confirm a lack of bone formation and healing response, consistent with a non-union (4 weeks following fracture surgery; Figures 1B,C). Dosing, withdrawal, and revision surgery timepoints were based on the previous work in this mouse line and cover the established bridging window in the control mice (Garcia et al., 2013; Liu et al., 2017; Hixon et al., 2021).

In the preparation for scaffold implantation, the subcutaneous adipose tissue was dissected from 12-week-old WT mice immediately following euthanasia (CO₂ asphyxiation). The dissected tissue was minced and placed in a digestion solution (0.1% collagenase and 1% penicillin/streptomycin in Dulbecco’s Modified Eagle’s Medium (DMEM)) at 37°C for 1 h. The samples underwent gentle inversion every 10 min. Following incubation, the tissue was transferred to a 10 cm culture dish, mechanically shredded with forceps, and then placed in an incubator for 30 min (37°C and 5% CO₂). At this time, the cell suspension was passed through a 70-µm sterile filter and centrifuged at 1300 rpm for 10 min at 4°C. The supernatant was removed and the cell pellet was resuspended in the growth medium (10% fetal bovine serum (FBS) and 1% penicillin/streptomycin in DMEM). All the cells were expanded to passage four which has previously been demonstrated to exhibit greater than 90% homogeneity for the MSC markers (CD13, CD29, CD44, CD73, and CD90) (Mitchell et al., 2006; Baer and Geiger, 2012; Dunham et al., 2020). All the adipose-derived stem cells (ADSCs) were immediately seeded onto the cryogel scaffolds at passage four for the in vivo implantation.

The cryogel scaffolds were fabricated as described previously in the Scaffold Fabrication. Three scaffold groups were prepared for implantation: scaffold only (acellular; no lentivirus), scaffold + ADSCs transduced to express GFP, and scaffold + ADSCs transduced to express BMP-2 (n = 5–7/group/timepoint). In the preparation for implantation, all the scaffolds were seeded with ADSCs (± lentiviral transduction). Briefly, for the lentiviral transduction, BMP-2 or GFP lentiviruses were immobilized on the scaffold and 1 million cells were seeded and transduced, as described previously in Scaffold Mediated Lentiviral Transduction. All samples were cultured for 6 days to allow for cell infiltration and proliferation to occur throughout the cryogels.

At 4 weeks following the fracture, the achievement of an atrophic non-union in Col1-tk mice was confirmed via the radiograph. All the mice were given buprenorphine SR (1 mg/kg, s. c.) for analgesia 1 h before the surgery. An 8- to 10-mm incision was created along the previously fractured femur and the muscle was gently separated from the non-union site, exposing 7 mm of the femur. The assigned scaffold type (n = 5–7/group/timepoint) was placed under the femur and wrapped around the non-union such that the surrounding muscle held the scaffold in place. The muscle was closed with 4.0 vicryl suture, then the incision was closed using the running subcuticular 6-0 PDS sutures (Ethicon). Immediately following the surgery, radiographs were used to confirm that the non-union site remained. Note that the controls included a sham group where the surgery was completed, but no scaffold was implanted. All the animals received dox water (1 μg/ml in 5% sugar water; Sigma) for the duration of the study, changed weekly. The animals were monitored radiographically for 2 or 8 weeks after the surgery, chosen from the previous studies (Douglas et al., 2012; Hao et al., 2016; Orth et al., 2017), to observe the potential rescue of the non-union site. Following euthanasia, both the right and left femurs were dissected for micro-CT and processed for the histological evaluation (Figures 1B,C).

Lateral radiographs were taken weekly following the full fracture until euthanasia (×3 magnification, Faxitron UltraFocus100; n = 5–7/group/timepoint). All the radiographs were blindly scored for the degree of healing using a modified RUST scoring system with both the sides of the fracture scored individually and a max score of 6 total: score = 1 designates only the fracture line is present, score = 2 is visible callus and fracture line, and score = 3 is completely bridged with no fracture visible (Cunningham et al., 2017; Fiset et al., 2018; Thomas and Kehoe, 2021). The mice were excluded from the study if there was a loss of fixation during healing or an atrophic non-union was not achieved by week 4 post-fracture (5 mice total).

Following euthanasia, the right and left femurs (n = 5–7/group/timepoint) were dissected and immediately fixed in 10% NBF for 24 h. After fixation, all the pins were removed from the bone and ex vivo scans were completed using micro-CT on the right and left femurs (VivaCT 40, Scanco Medical AG, Switzerland; 10.5 μm voxel size, 55 kV, 145 μA, 300 ms integration time). For the right fractured femurs, a 600 slice (6.3 mm length and diameter) ROI cylinder was centered at the fracture midpoint and thresholds of 180 and 250 per mile was used for 2 and 8 weeks timepoints, respectively, to segment the total bone (callus + cortical). A high-density threshold of 460 per mile was then used to segment out the original cortical bone of all the right (fractured) femurs. The cortical bone volume was subtracted from the total bone volume of each sample to determine the volume of new callus bone formed. In some samples, the fracture lines were no longer visible due to healing, so the point of fracture was found from the radiographic images at 0–2 weeks post-fracture (femoral head to fracture midpoint distance). For the left (intact) femurs, a 100 slice (1.05 mm) ROI was analyzed to compare the systemic changes due to the surgery or BMP-2 production. Following scanning, all the femurs were decalcified and processed for histology.

Following micro-CT, the samples were decalcified in 14% EDTA (pH 7.0) for 2 weeks (n = 5–7/group/timepoint). Standard paraffin processing was completed where all the fractured femurs were sectioned longitudinally at 5 µm thickness. Slides from all the samples were stained with picrosirius red/alcian blue to visualize the different tissues (mineralized callus, cartilage, and fibrous tissue/other) at the fracture site. Immunohistochemistry (IHC) was also used to localize the endomucin expression to evaluate the presence of vasculature throughout the scaffold, as well as the HSV-tk and GFP expression to identify the role of the tk+ and GFP+ cells in healing, respectively. To accomplish this, the paraffin slides were first deparaffinized in xylene and rehydrated in graded ethanol. Next, incubation with 3% H₂0₂ for 5 min blocked the endogenous peroxidases. Note that proteinase K was used for the antigen retrieval for endomucin slides prior to blocking (5 min, 23°C). HSV-tk and GFP endogenous epitopes were blocked with 10% goat serum (abcam—ab7481) in PBS at RT for 1 h. The sections were then incubated with rabbit polyclonal anti-HSV tk (1:100; gift from William Summers, Yale) and rabbit-anti-GFP (1:500; Life Technologies A11122) overnight at 4°C, followed by a secondary polymer-HRP anti-rabbit antibody incubation for 1 h (1:200 Dako P0448). The endomucin staining followed the manufacturer’s instructions (Vectastain Elite ABC kit, Rat IgG PK-6104) and used the primary antibodies for endomucin (1:400, eBioV.7C7 #14–5851–85). All the IHC samples were incubated with DAB chromagen (brown; Vector DAB, SK-4105, 30–45 s) and counterstained with modified Mayer’s hematoxylin (Vector Hematoxylin QS, H-3404). Images were obtained on a NanoZoomer slide scanner (20x; Hamamatsu Photonics). Picrosirius Red/alcian blue, HSV-tk, and endomucin images were blinded and qualitatively assessed for the callus and scaffold composition.

Prior to the experiments, sample sizes for the study were calculated using a power analysis with α = 0.05 and β= 0.20 (https://www.stat.ubc.ca/∼rollin/stats/ssize/n2.html). The estimates of sample variance and effect size were based on the previous data and biological importance (McKenzie et al., 2019). One-way ANOVA with the Tukey post hoc test was used for in vitro bone volume and bone mineral density analysis. A chi-square test was used to assess the fracture union based on the RUST radiographic scoring at 2 and 8 weeks post-scaffold implantation. Two-way ANOVA with the Tukey post hoc test was used to correct for multiple comparisons for ex vivo micro-CT. Significance was considered at p-values < 0.05, with trends noted at p < 0.10. All the data analysis was performed using Prism (version 9; GraphPad Software, La Jolla, CA, USA).

The ADSCs were seeded onto the cryogel scaffolds were cultured for 6 days and then cryosectioned to visualize cell adhesion and infiltration (Figure 2A). As shown by DAPI stain (blue), the ADSCs were able to infiltrate the matrix of all the cryogel scaffolds in the short window of culture, consistent with prior results with other cell types (Hixon et al., 2016; Hixon et al., 2017a). To verify the BMP-2 production, following an initial 6 days of culture, the cells were given osteogenic medium ± the addition of doxycycline (dox; Figure 2B). The non-transduced and transduced cells that did not receive dox (no dox) had negligible BMP-2 production over 28 days of culture. Similarly, a low dose of 0.1 μg/ml of dox was not sufficient to initiate the detectable BMP-2 production. However, when the transduced cells were exposed to 1 μg/ml of dox, a significantly higher production of BMP-2 was noted in all the samples which lasted over 28 days. It should be noted that while there was some variability in the BMP-2 secretion, these are the biological replicates. Thus, the seeded constructs may have been of slightly different sizes, seeded with slightly different numbers of cells, or received slightly different amounts of virus. Following the culture, micro-CT was used to quantify the mineralized volume and density (Figures 2C,D). The transduced cells given 1 μg/ml of dox and thus, producing BMP-2, had significantly more mineralized volume than the non-transduced cells and significantly higher bone mineral density than the acellular scaffolds (p < 0.05). Additionally, the transduced cells either left untreated or supplemented with 0.1 μg/ml of dox also had significantly higher bone mineral density than the acellular scaffolds (p < 0.05). Note that due to the presence of osteogenic medium, some mineralized volume and bone mineral density levels, when the cells are present, are not unexpected. These results indicate that cryogel scaffolds support mineralization by the ADSCs in vitro, and that the mineralized volume is enhanced by the dox-inducible cellular production of BMP-2.

Blinded scoring of the radiographs was completed weekly until sacrifice at 2 or 8 weeks post-scaffold implantation (6 or 12 weeks post-fracture; Figure 3). Less than 50% of the animals treated with either scaffold only or with a sham implantation procedure showed complete bridging of both the cortices at the fracture site at 12 weeks. As both the groups consisted of Col1-tk animals treated with GCV, they were not expected to heal. In comparison, the scaffolds seeded with GFP+ cells and the scaffolds seeded with cells producing BMP-2 were the only groups that reached the complete bridging in all the animals by 12 weeks. Specifically, the fractures in all the animals treated with GFP+ scaffolds had complete bridging (RUST score = 6) after 6 weeks, while over 70% of the BMP-2+ scaffolds were bridged by week 8 and 100% by week 12. Notably, in all the animals treated with the BMP-2+ scaffolds, mineralization was observed radiographically by week 5–7 (1–3 weeks following implantation) throughout the implanted scaffold surrounding the fracture site. In comparison, the GFP+ groups only exhibited mineral within the scaffold in 2 of the 14 samples at varying timepoints. There was no significant difference between the radiograph scores of any groups at 6 weeks. However, the radiograph scores of both the scaffold and sham groups were significantly different than the radiograph scores of both the GFP and BMP-2 groups at 12 weeks. Note that there was no significance between the sham and scaffold groups or the GFP and BMP-2 groups at this timepoint (χ2, p < 0.05). In summary, the cellularized scaffolds promoted the radiographic healing, and the addition of BMP-2 stimulated the scaffold mineral formation.

Following the treatment with scaffolds for 2 or 8 weeks, the animals were sacrificed for micro-CT and histological evaluation. Ex vivo micro-CT analysis determined the callus bone volume at both 2 and 8 weeks for all the groups (Figure 4). Note that the 2 and 8 weeks timepoints had a threshold at 180 and 250 per mile, respectively, to isolate the scaffold mineral and thus, cannot be directly compared. At 2 weeks, the BMP-2+ cellularized scaffolds had significantly more bone volume than the acellular scaffold group (p < 0.05) and also trended higher compared to the sham control (p < 0.0668; Figure 4A). Three-dimensional reconstructions further demonstrate this, with the mineral formation around the non-union site and infiltrating into the surrounding scaffold. At 8 weeks, the BMP-2+ cellularized scaffolds had significantly more bone volume than all the other groups (p < 0.001; Figure 4B). Three-dimensional reconstructions at this timepoint also demonstrate mineral formation around the non-union site within the surrounding scaffold. Additionally, while the fracture line remains apparent between 2 and 8 weeks in the majority of the scaffold only and sham groups, the GFP+ and BMP-2+ cellularized scaffolds exhibited a higher ratio of bridging and remodeling of the cortical bone between 2 and 8 weeks. We assessed the skeletal phenotype of the left (intact) femora using micro-CT and found no significance between any groups (p < 0.05; Supplementary Table S1), indicating no systemic differences between the treatments. In summary, the cellularized scaffolds promoted bridging at the fracture site, but only BMP-2+ scaffolds produced a significant new bone surrounding the fracture site.

Following the micro-CT analysis of 2-8 weeks samples, histological evaluation was completed using picrosirius red/alcian blue stain (Figure 5). At 2 weeks post-sham surgery (6 weeks post-fracture), the sham control group displayed minimal woven bone formation and high levels of fibrous tissue. After 8 weeks, incomplete healing was still noted in 5 of the 6 samples, despite woven bone formation and some cortical bridging at all of the fracture sites. Similarly, the scaffold only treatment displayed some calcified areas with almost no cartilage present and minimal infiltration into the surrounding scaffold, showing incomplete endochondral bone formation. After 8 weeks, full bridging (denoted by continuous cortical bone) had visually occurred in 3 of the 7 samples, with the remaining samples possessing fibrous tissue and minimal cartilage. In the animals treated with the GFP+ cellularized scaffolds, new bone was formed at 2 weeks with the presence of minimal cartilage and fibrous tissue. By 8 weeks postimplantation, full bridging was achieved in 7 of the 7 samples, with the new bone infiltrated into the surrounding scaffold and no fibrous tissue or cartilage. When the cryogel scaffolds seeded with the BMP-2-expressing cells were implanted, cartilage was visualized surrounding the non-union site and within the surrounding scaffold in 6 of the 7 samples at 2 weeks. After 8 weeks, bridging was visually achieved in all the samples with the mineral infiltrating any surrounding scaffold still present; no fibrous or cartilage callus tissue was observed. In summary, the majority of samples in the sham and scaffold only groups displayed remnants of the original fragmented cortical bone with visible fracture surfaces and lacked complete bridging, whereas the GFP+ and BMP-2+ groups displayed continuous cortical surfaces such that the fracture line was no longer visible in most samples.

The sections were also stained for endomucin to identify the presence of vasculature, HSV-tk to localize the tk + cells, and GFP to discern the role of GFP+ cells in both the endogenous callus at the non-union site, as well as within the scaffold (Figure 6, Supplementary Figures S1, S2). Overall, all the groups possessed a large number of vessels directly at the fracture site and closely associated with the woven bone formation at 2 weeks (Figure 6A). After 8 weeks, the vasculature at this site was greatly reduced with little to no endomucin staining in all the groups. When the surrounding scaffolds were examined (Figure 6B), the vessels were visible both within the scaffold matrix and accompanying the new bone formation at 2 and 8 weeks in all the groups treated with a scaffold ± cells. The new bone was typically formed on the edges of the scaffold, both proximal and distal to the fracture site. With respect to the HSV-tk staining, the positively stained osteoblasts and osteocytes were detected within the bone tissue in some samples from every group at both 2 and 8 weeks (Supplementary Figure S1). After 8 weeks, all the BMP-2+ scaffolds also possessed HSV-tk + cells (red arrows) localized to the new bone formed within the scaffold, indicating the invasion of host cells. Finally, GFP+ staining was found within the scaffold around the fracture site at 2 weeks in all the samples treated with scaffolds seeded with GFP+ cells (Supplementary Figure S2). However, after 8 weeks, there was no GFP+ staining in 5 of the 7 samples, with very little staining in the remaining 2 samples, demonstrating that the GFP+ cells were no longer present at 8 weeks in the majority of samples.