Periosteal Mesenchymal Progenitor Dysfunction and Extraskeletally-Derived Fibrosis Contribute to Atrophic Fracture Nonunion

Abstract

Atrophic nonunion represents an extremely challenging clinical dilemma for both physicians and fracture patients alike, but its underlying mechanisms are still largely unknown. Here, we established a mouse model that recapitulates clinical atrophic nonunion through the administration of focal radiation to the long bone midshaft 2 weeks before a closed, semistabilized, transverse fracture. Strikingly, fractures in previously irradiated bone showed no bony bridging with a 100% nonunion rate. Radiation triggered distinct repair responses, separated by the fracture line: a less robust callus formation at the proximal side (close to the knee) and bony atrophy at the distal side (close to the ankle) characterized by sustained fibrotic cells and type I collagen-rich matrix. These fibrotic cells, similar to human nonunion samples, lacked osteogenic and chondrogenic differentiation and exhibited impaired blood vessel infiltration. Mechanistically, focal radiation reduced the numbers of periosteal mesenchymal progenitors and blood vessels and blunted injury-induced proliferation of mesenchymal progenitors shortly after fracture, with greater damage particularly at the distal side. In culture, radiation drastically suppressed proliferation of periosteal mesenchymal progenitors. Radiation did not affect hypoxia-induced periosteal cell chondrogenesis but greatly reduced osteogenic differentiation. Lineage tracing using multiple reporter mouse models revealed that mesenchymal progenitors within the bone marrow or along the periosteal bone surface did not contribute to nonunion fibrosis. Therefore, we conclude that atrophic nonunion fractures are caused by severe damage to the periosteal mesenchymal progenitors and are accompanied by an extraskeletal, fibro-cellular response. In addition, we present this radiation-induced periosteal damage model as a new, clinically relevant tool to study the biologic basis of therapies for atrophic nonunion.

Introduction

Unlike many other tissues that heal with a scar, bone is unique in that it can repeatedly and routinely heal traumatic or acute injury through regeneration in a spatiotem-porally predictable fashion.⁽¹⁾ However, the healing process is not completely perfect and still occasionally fails, resulting in a 5% to 10% rate of delayed healing.⁽²⁾ Among the 1 million delayed healing cases per year in the US alone, one-sixth (more than 150,000 per year) will progress to nonunion, placing a substantial physical, medical, emotional, and financial burden upon affected individuals and our society as a whole. Types of nonunion fractures include hypertrophic, oligotrophic, and atrophic forms.⁽³⁾ Hypertrophic and oligotrophic nonunions have nonbridging calluses that contain cartilage. Originating from inadequate fracture stability, they are normally treated by a stabilizing osteosynthesis with great success. In contrast, atrophic nonunions have minimal callus or cartilage and instead form fibrotic tissue within the fracture gap. They represent a substantial treatment challenge for surgeons, medical morbidity to the patient, and a biocellular conundrum for the clinical investigator. Therefore, development and understanding of a nonunion model is crucial to solve this problem.

Mouse is a common species of choice for mechanistic orthopedic research because of the availability of genetic tools. Recent studies in mouse models engineered for lineage tracing demonstrated that mesenchymal progenitors, labeled by Prrx1-CreER,⁽⁴⁾ αSMA-CreER,⁽⁵⁾ or GlH-CreER,⁽⁶⁾ are the main contributors to callus cells during fracture healing. Nevertheless, the study of fracture nonunion is limited by the intrinsically high bone regeneration ability in rodents, which exhibit near 100% fracture healing rates, even without surgical intervention. To date, segmental defect and surgical periosteum resection/ destruction are the only methods to generate fractures that progress to nonunion.⁽⁷⁾ However, while these models serve as interesting assays for overall bone regeneration, they do not reflect most clinical nonunion cases in which fractured bones are placed in close proximity and which commonly arise from trauma resulting from application of energy rather than surgical precision.

Radiation therapy is a common treatment for cancer but often results in collateral damage on bone within the irradiated field. One of the late side effects of radiotherapy is fracture. Unfortunately, fractures occurring after radiation are difficult to treat and are associated with very high rates of delayed union and nonunion.⁽⁸⁾ The fracture healing time in patients with soft tissue tumors exceeds a year and is compounded by a 45% nonunion rate.⁽⁹⁾ This clinical relationship between radiation therapy, fracture, and healing provided the impetus for utilizing radiation as a model to induce and study atrophy during bone regeneration. Previously, we used a focal irradiator (Small Animal Radiation Research Platform [SARRP]), which can mimic clinical focal radiation therapy in rodents with submillimeter accuracy, to delineate the mechanism of radiation damage on metaphyseal trabecular bone and to investigate potential treatments for this devastating condition.^(10–12) Here, we extended our research to traumatic bone fracture and established a highly reproducible and clinically relevant mouse nonunion model. Characterization of this model revealed the underlying cellular and molecular mechanisms of nonunion formation. Furthermore, we took advantage of multiple lineage tracing approaches to delineate the cellular origin of fibrotic tissue that blocks bony bridging in nonunions.

Materials and Methods

Animal study design

Specific pathogen-free male 2-month-old C57BL/6 mice were purchased from the Jackson Laboratory (Bar Harbor, ME, USA). Col2-Cre Rosa-tdTomato (Col2/Tomato) and Gli1-CreER Rosa-tdTomato (Gli1/Tomato) mice were generated by breeding Rosa-tdTomato (Jackson Laboratory) mice with Col2-Cre⁽¹³⁾ (Jackson Laboratory) and Gli1-CreER⁽¹⁴⁾ (Jackson Laboratory) mice, respectively. αSMA-CreER Rosa-tdTomato (αSMA/Tomato) was a gift from Dr Ivo Kalajezc.⁽⁵⁾ In accordance with the standards for animal housing, mice were group housed at 23°C to 25°C with a 12-hour light/dark cycle and allowed free access to water and standard laboratory pellets. To induce CreER activity, mice received tamoxifen injections (75 mg/kg/d) either for 2 days starting from the day before fracture (αSMA/Tomato mice) or for 5 days starting from 5 days before fracture (Gli1/Tomato mice). A clinically relevant radiation dose of 2 × 8 Gy on day 1 and day 3 was delivered to the midshaft of the right tibias (5 mm in diameter) from a focal irradiator (SARRP, Xstrahl, Suwanee, GA, USA) at a rate of 1.65 Gy/min with the aid of built-in micro-computed tomography (microCT) and X-ray as described previously.⁽¹⁵⁾ Two weeks later, closed, transverse fractures were made within the irradiated area and the same area in the contralateral legs via a blunt guillotine with a preinserted intramedullary pin. Bones were harvested at various time points for microCT, histology, and 3-point bending assays (Fig. 1A). In samples that received periosteum resection instead of radiation to induce nonunion, the anterior side of the tibia was minimally exposed immediately after fracture and the periosteum 3 mm proximal and/or distal to the fracture line was removed with a surgical scalpel. The skin was then closed with sutures.

Fig. 1.

Prior focal radiation has distinct effects on callus formation at two sides of the fracture line and results in 100% nonunion. (A) Schematic design of animal experiments. R = right tibia; L = left tibia; Tam = tamoxifen injections; CT = microCT scanning; H = histology. (B) Representative 2D microCT images of fracture calluses at 1, 2, 4, and 6 weeks postfracture. Calluses were formed at both proximal (pro) and distal (dis) sides in nonirradiated (NR) bones and formed only at the proximal side in prior irradiated (R) bones. Dashed red lines depict the fracture lines. Scale bar = 1 mm. (C) 3D reconstructed tibias at 6 weeks postfracture show nonunion in bones receiving prior radiation. A red arrow points to the site of nonunion. Scale bar = 1 mm. (D) Callus tissue volume (TV), bone volume (BV), and bone volume fraction (BV/TV) of fracture calluses at 1, 2, 4, and 6 weeks postfracture were measured. The fracture callus was divided into proximal and distal sides for quantification. (E) Fracture healing scores were quantified on tibias at 6 weeks postfracture. (F) Three-point bending test was performed on tibias at 6 weeks postfracture. Values are mean ± SD. n = 7–11 mice/time point. ^ap < 0.05; ^bp < 0.01; ^cp < 0.001 R versus NR; ^*p < 0.05; ^$p < 0.01; ^#p < 0.001 distal versus proximal. Paired Student’s t test.

MicroCT analysis

Tibias harvested at 1, 2, 4, and 6 weeks postfracture were scanned at the fracture sites using a VivaCT 40 (Scanco Medical AG, Brüttisellen, Switzerland) at a 10.5-μm isotropic voxel size to acquire a total of 686 microCT slices centering around the fracture site. A user-defined contouring method was used to define the callus perimeter. This semiautomated segmentation method analyzes the callus outside the preexisting cortical bone. All images were first smoothed by a Gaussian filter (sigma = 1.2, support = 2.0) and then applied with a threshold (240) corresponding to 30% of the maximum available range of image grayscale values to distinguish mineralized tissue from unmineralized and poorly mineralized tissue. The callus region surrounding the cortical bone was contoured for trabecular bone analysis. 3D standard microstructural analysis was performed to determine total callus volume (TV), bone volume (BV),and bone volume fraction (BV/TV) at the proximal and distal sides of fracture. Callus tissue area (TA) and bone area (BA) of 200 cross-sectional slices proximal to and 200 slices distal to the fracture line were calculated individually. Based on microCT images, 6-week fracture samples were assigned fracture healing scores according to a modified five-point radiographic scoring system.⁽¹⁶⁾ To quantify vessel volume in the callus, mice were perfused with microfil (Flow Tech Inc., Carver, MA, USA; Microfil MV-122) at euthanized. Fractured tibias were harvested and scanned by microCT at a 10.5-μm isotropic voxel size before and after 3 weeks of decalcification by 10% EDTA. A total of 686 slices, either at the distal side or at the proximal side of fracture line, were used to determine the vessel volume.

Mechanical testing

Tibias harvested at 6 weeks after fracture were placed on a 3-point bending fixture and loaded with mechanical force at the previously fractured site using an Instron 5542 (Instron, Norwood, MA, USA) with a rate of loading of 1.8mm/min, a loading tip radius of 0.3 mm, a support points span of 10 mm, and the direction of loading perpendicular to the sample. The load-to-failure curve was recorded, and peak load, stiffness, and energy to failure were quantified.

Histology and immunohistochemistry (IHC)

Fractured tibias were fixed in 4% PFA, decalcified in 10% EDTA for 3 weeks, and processed for paraffin embedding. A series of 6-μm-thick longitudinal sections were cut across the entire fracture callus from one side of cortical bone to the other. For each bone, a central section with the largest callus area, as well as two sections at 192 μm (~¼ bone width) before and after the central section were stained with Safranin-O/Fast green and quantified for cartilage area, bone area, and fibrosis area by ImageJ. Paraffin sections adjacent to the central sections were used for picrosirius red staining and IHC. After antigen retrieval, slides were incubated with primary antibodies, including rabbit anti-osteocalcin (Takara Clontech, Mountain View, CA, USA; m173), rabbit anti-VEGF (Abcam, Cambridge, MA, USA; AB46154), goat anti-Osterix (Santa Cruz Biotechnology, Dallas, TX, USA; sc-22538), rabbit anti-type II Collagen (Abcam, AB34712), goat anti-Sox9 (R&D Systems, Minneapolis, MN, USA; AF3075), and rat anti-Endomucin (Santa Cruz Biotechnology, sc-65495), at 4°C overnight, followed by binding with biotinylated secondary antibodies and DAB color development. For TRAP staining, tartrate-resistant acid phosphatase (TRAP) assay kit (Sigma-Aldrich, St. Louis, MO, USA; 387A-1KT) was used.

To obtain frozen sections for immunofluorescent imaging, fractured tibias were fixed in 4% PFA for 1 day, transferred to 30% sucrose in PBS overnight, and embedded in OCT compound (ThermoFisher Scientific, Waltham, MA, USA) for frozen sectioning at a thickness of 6 μm, aided by the use of cryofilm 2C (SECTION-LAB Co. Ltd., Hiroshima, Japan). After glued to slides, sections were incubated with rat anti-Endomucin at 4°-C overnight followed by Alexa Fluor 488-conjugated goat anti-Rat IgG secondary antibody (Abcam, ab150157). For EdU staining, mice received 1.6 mg/kg EdU 3 hours before death and the staining was carried out according to the manufacturer’s instructions (Click-iT^™ EdU Alexa Fluor^™ 647 Imaging Kit, ThermoFisher Scientific, c10340).

Human nonunion scar tissue samples

Nonunion scar tissues were prepared from the de-identified specimens obtained from the open reduction and internal fixation surgeries of patients with long bone nonunion fractures (n = 9). The diagnosis of nonunion is at least 1 year’s duration according to radiological proof. Samples were fixed in 4% PFA overnight, decalcified in 20% EDTA for 4 to 12 weeks, followed by paraffin embedding and staining with H&E, picrosirius, and Safranin O/Fast green. For IHC of osteocalcin, slides were incubated with anti-human osteocalcin antibody (Abcam, AB198228) at 4°C overnight, followed by binding with biotinylated secondary antibody and DAB color development.

Periosteal mesenchymal progenitor culture

Intact, age-matched mouse tibias were dissected free of surrounding tissues and both ends were removed at the growth plate sites. The remaining bone fragments were digested in 2 mg/mL collagenase A and 2.5 mg/mL trypsin. Cells from the first 3 minutes of digestion were discarded and cells from a subsequent 20 minutes of digestion were seeded in the growth medium (αMEM supplemented with 15% fetal bovine serum [FBS] plus 55 μM β-mercaptoethanol, 2mM glutamine, 100IU/ mL penicillin and 100 μg/mL streptomycin) for periosteal mesenchymal progenitor culture. For colony-forming unit fibroblast (CFU-F) assay, cells were seeded at 0.3 × 10⁶ cells/ T25 flask. Seven days later, flasks were stained with 3% crystal violet to quantify CFU-F numbers. To study the radiation effect on progenitor proliferation, cells were seeded at 5600cells/cm², irradiated (8Gy) the next day, and counted at 0, 1, 2, 4, and 5 days postradiation. To study the radiation effect on differentiation, cells were irradiated (8 Gy) when confluent, then switched to either osteogenic medium (αMEM with 10% FBS, 10 nM dexamethasone, 10 mM β-glycerophosphate, 50 μg/ mL ascorbic acid, 100IU/mL penicillin, and 100 μg/mL streptomycin) for 2 weeks followed by alizarin staining, or adipogenic medium (αMEM with 10% FBS, 0.5 mM isobutylmethylxanthine, 10 mM indomethacin, 1 μM dexamethasone, 10 μg/mL insulin, 100IU/mL penicillin, and 100 μg/mL streptomycin) for 1 week followed by Oil Red O staining. For chondrogenic differentiation, after radiation progenitor cells were suspended at 1.25 × 10⁶ cells/mL in chondrogenic medium (high-glucose DMEM, 100 μg/mL sodium pyruvate, 1% ITS+Premix, 50 μg/mL ascor-bate-2-phosphate, 40 μg/mL L-proline, 0.1 mM dexamethasone, 10 ng/mL TGF-β3,100 IU/mL penicillin, and 100 μg/mL streptomycin). Aliquots of 200 μL cell solution were then distributed to a V-bottomed 96-well plate followed by centrifugation at 300G for 5 minutes. Pellets were cultured for 3 weeks and then harvested for paraffin sections with Alcian blue staining. To culture cells under hypoxia condition, cell culture plates were placed in a Modular Incubator Chamber (Billups-Rothenberg, Del Mar, CA, USA) supplied with 0.1% oxygen.

RNA analysis of marker genes

To quantify the expression level of marker genes, cells were collected in Tri Reagent (Sigma, St. Louis, MO, USA) for RNA purification. A Taqman Reverse Transcription Kit (Applied BioSystems, Inc., Foster City, CA, USA) was used to reverse-transcribe mRNA into cDNA. After this, quantitative real-time PCR (qRT-PCR) was performed using a Power SYBR Green PCR Master Mix Kit (Applied BioSystems, Inc). The primer sequences for the genes used in this study are listed in Supplemental Table S1.

Statistics

Data are expressed as means ± standard deviation (SD) and analyzed by paired Student’s t test for comparison of irradiated and nonirradiated contralateral bones or comparison of the proximal and distal sides of fracture, and by unpaired Student’s t test for comparison of cell culture samples followed by Bonferroni adjustment for multiple comparisons using Prism 5 software (GraphPad Software, San Diego, CA, USA). For cell culture experiments, observations were repeated independently at least three times with a similar conclusion, and only data from a representative experiment are presented. Values of p < 0.05 were considered statistically significant.

Study approval

All animal experiments were performed in compliance with the NIH’s Guide for the Care and Use of Laboratory Animals (National Academies Press, 2011) and under protocols approved by the University of Pennsylvania Institutional Animal Care and Use Committee, Philadelphia, PA, USA.

Results

Prior focal radiation causes fracture nonunion in mouse long bones

To investigate the effect of prior radiation on fracture healing, we irradiated the central midshaft of the mouse right tibia using a SARRP focal irradiator and, 2 weeks later, generated a closed, semistabilized, and transverse fracture within the irradiated area, as well as in the corresponding site in the contralateral, nonirradiated leg (Fig. 1A). Radiation at doses used in our experiments rapidly caused significant damage to bone marrow cells, leading to drastically decreased cellularity and dilated vasculature (Supplemental Fig. S1). However, this bone marrow damage seemed to be reversed by the time immediately preceding fracture.

In nonirradiated tibias, weekly microCT scans showed that callus formation reached a peakat 2 weeks,gradually reduced its size at 4 weeks, and achieved 100% bone bridging and healing at 6 weeks postfracture (Fig. 1B–D). In contrast, calluses in irradiated tibias were much smaller in size and never consolidated into bone bridging at 6 weeks, resulting in 100% rate of nonunion (Fig. 1B–D) and a significantly reduced healing score (Fig. 1E). Careful examination of microCT images revealed distinct repair processes at the distal side (the region closes to the ankle) compared with the proximal side (the region closes to the knee) of fractures in irradiated bones. Although the proximal side formed a bone-containing callus, albeit with reduced TV and BV compared with those at the corresponding site of nonirradiated bone fracture, we did not detect any callus or bone formation at the distal side (Fig. 1B, D; Supplemental Fig. S2). By week 6 postfracture, the proximal callus in irradiated bone grew past the fracture gap toward the distal site but failed to adherently bridge the fracture gap (Fig. 1B). Mechanical testing of fractured bones at 6 weeks showed that previously irradiated tibias had 69%, 75%, and 69% decreased peak load, stiffness, and energy to failure, respectively (Fig. 1F). These data confirm that bone regeneration was significantly impeded in irradiated bones.

Fibrotic tissue replaced soft callus formation at the distal side of the fracture after radiation

Next, we performed histology to investigate the underlying cellular mechanism behind this impaired fracture healing. As shown by Safranin O/Fast green staining in Fig. 2A, B, nonirradiated bones exhibited normal fracture repair processes, with the regions near the fracture gap undergoing endochondral ossification (cartilage formation followed by bone replacement) and the regions distant from the fracture gap undergoing intramembranous ossification (direct bone formation). Surprisingly, although the proximal side of irradiated bones showed similar, though reduced, cartilage and bone structures in the callus, the distal side consisted of only fibroblast-like cells surrounding the cortical bones from week 1 to 6 after fracture. Picrosirius red staining showed that these cells express abundant type I collagen (Fig. 2C), indicative of fibrotic tissue. Quantification confirmed that the proximal callus area, cartilage area, and bone area were all significantly reduced by 63%, 47%, and 68%, respectively, at 1 week after fracture and by 40%, 7%, and 34%, respectively, at 2 weeks after fracture in irradiated bones compared with contralateral bone fractures (Fig. 2D). Note that the proximal side of irradiated bones showed both a reduced callus size and a delayed repair process, with cartilage remaining at 4 weeks, a time point when cartilage had already been remodeled into bone during normal fracture repair. The fibrotic tissue area appeared at 1 week postfracture, reached maximum volume at 2 weeks, and persisted afterward (Fig. 2E). This radiation effect was not unique to the tibia. Similar radiation at the central midshaft of mouse femurs followed by a fracture 2 weeks later generated the same 100% nonunion, with only type I collagenabundant fibrotic tissue observed at the distal side of fracture (Supplemental Fig. S3).

Fig. 2.

Fibrotic tissue is formed at the distal side of fractures in the prior irradiated tibias. (A) Representative Safranin O/Fast green staining images of fracture calluses at 1, 2,4, and 6 weeks postfracture. Cartilage (labeled as 2 and 3) and woven bone (labeled as 1 and 4) were observed at both proximal (pro) and distal (dis) sides in nonirradiated (NR) fractures but only at the proximal side in prior irradiated (R) fractures. Dashed black lines depict the fracture lines. Fibrotic tissues outlined in black curves were observed at the distal side. Scale bar = 1 mm. (B) Magnified images of regions 1, 2,3, and 4 at 2 weeks postfracture from A. Scale bar = 100 μm. (C) Picrosirius red-stained images of the same regions show that fibrotic tissue (regions 3 and 4) contains abundant type I collagen matrix. Scale bar = 100 μm. (D) Callus area, cartilage area, and bone area were measured at the proximal and distal sides of fracture in nonirradiated and irradiated tibias at 1,2, and 4 weeks postfracture. (E) Fibrosis area was measured at the distal side of fractures. Values are mean ± SD. n = 7–11 mice/time point. ^ap < 0.05; ^bp < 0.01; ^cp < 0.001 R versus NR; ^*p < 0.05; ^$p < 0.01; ^#p < 0.001 distal versus proximal. Paired Student’s t test.

Fibrotic tissue lacks differentiation ability and vessel infiltration, mimicking human nonunion samples

To characterize this fibrotic tissue, we stained fracture sections with osteogenic and chondrogenic markers. At 2 weeks postfracture, while newly formed woven bone area at both the distal and proximal sides of nonirradiated fractures and at the proximal side of irradiated fractures had many Osterix⁺ and Osteocalcin⁺ cells, fibrous cells at the distal side of irradiated fractures were all negative for these osteogenic markers (Fig. 3A). Similarly, fibrous cells did not stain for the chondrogenic markers type II collagen and Sox9, in sharp contrast to the cartilage tissue in normal fracture healing (Fig. 3B).

Fig. 3.

Fibrotic cells lack osteogenic and chondrogenic differentiation ability. (A) IHC of osteogenic (Osterix, Osteocalcin) markers at the proximal and distal sides of the callus in nonirradiated (NR) and irradiated (R) bones at 2 weeks after fracture. CB = cortical bone; WB = woven bone; FT = fibrotic tissue. Arrows point to positive cells. Scale bar = 100 μm. (B) IHC of chondrogenic (type II collagen, Sox9) markers at the proximal and distal sides of callus at 2 weeks after fracture. C = cartilage. Scale bar = 50 μm.

Normal callus contains abundant vasculature that is essential for fracture healing. However, at 2 and 4 weeks postfracture, fibrotic tissue at the distal side did not show any vessel staining (Fig. 4A). Microfil infiltration, followed by microCT scanning, revealed an 85% decrease in vessel volume (VV) at the distal site of irradiated fractures compared with that in nonirradiated fracture (Fig. 4B, C). Further staining showed that these fibrous cells did not express the angiogenic factor vascular endothelial growth factor (VEGF), in contrast to a robust VEGF expression in normal calluses (Fig. 4D). Because no vessel invasion was detected in the fibrotic tissue, this led to the absence of resorbing osteoclasts, as identified by TRAP staining, in the area as well (Fig. 4E). Thus, we conclude that this fibrotic tissue formed in nonunion fracture after a prior radiation lacks osteogenic and chondrogenic differentiation abilities and is devoid of vessel invasion.

Fig. 4.

Fibrotic tissue is devoid of vessel infiltration. (A) IHCof Endomucin, an endothelial cell marker, at the proximal and distal sides of the callus at 2 and 4 weeks postfracture. NR = nonirradiated; R = irradiated; WB = woven bone; C = cartilage; FT = fibrotic tissue. Scale bar = 50 μm. (B) Representative microCT images of microfil perfusion at 2 weeks postfracture. MicroCT scans were performed before (bone and vessel) and after decalcification (vessel only) on the same bone. Dashed lines depict the fracture lines. Pro = proximal; dis = distal. Scale bar = 1 mm. (C) Quantification of vessel volume (VV) at the proximal (pro) and distal (dis) sides of the callus. Values are mean ± SD. n = 5 mice. ^ap < 0.05 R versus NR; ^*p < 0.05 dis versus pro. Paired Student’s t test. (D) IHC of VEGF at the proximal and distal sides of the callus at 2 weeks after fracture. Scale bar = 50 μm. (E) IHC of TRAP at the proximal and distal sides of the callus at 2 weeks after fracture. Scale bar = 50 μm.

During bone repair, bone marrow within the fracture region needs to be repaired, too. In nonirradiated bone, we observed that bone marrow cells at the distal side, but not those at the proximal side, developed a necrotic morphology shortly after fracture and gradually recovered at 4 weeks (Supplemental Fig. S4). Consistent with changes outside of bone, this bone marrow necrosis persisted in the irradiated bones and showed no sign of recovery at 4 weeks, indicating permanent radiation damage to the bone marrow.

To investigate the clinical relevance of this fibrotic tissue surrounding the cortical bone, we collected waste scar tissue surrounding the fracture ends of nonunion patients for staining. Cells in those tissues showed typical fibroblastic morphology with no bone, cartilage, or vessels detected (Supplemental Fig. 5A). Moreover, Picosirus red staining revealed abundant type I collagen matrix (Supplemental Fig. 5B). Further staining also showed no osteogenic marker (osteocalcin) expression (Supplemental Fig. 5C). Therefore, the fibrotic tissue in our mouse nonunion model parallels the pathology of human fracture nonunion samples.

Radiation damages periosteal mesenchymal progenitors and blunts their injury responses

Periosteal mesenchymal cells play a pivotal role in fracture healing.⁽¹⁷⁾ We and others recently established a mouse model (Col2-Cre Rosa-tdTomato, Col2/Tomato) to label the entire lineage of bone marrow mesenchymal cells, from stem cells to osteoblasts/osteocytes and adipocytes, inside mouse long bones.^(15,18) Interestingly, in this model, Tomato also marks the periosteum that continuously covers the cortical bone surface, with more abundance at the proximal region (Fig. 5A, B). Strikingly, at 2 weeks after radiation and right before fracture, Tomato⁺ cells within the radiation field were significantly reduced, whereas those in the neighboring area remained the same, suggesting that radiation locally damaged periosteal mesenchymal cells (Fig. 5B, C). Because blood supply is important for fracture healing, we also characterized the radiation effect on vessels along the periosteal bone surface. Similar to Tomato⁺ cells, tibial periosteal vessels were more abundant at the proximal region. Focal irradiation significantly decreased the number of periosteal vessels 2 weeks later and this effect was isolated to the irradiated field (Fig. 5B, D).

Fig. 5.

Prior radiation reduces the periosteum responses toward fracture. (A) H&E images of longitudinal tibial sections from Col2/Tomato mice indicating the focal radiation (R) field depicted by two dashed lines. Scale bar = 1 mm. (B) Fluorescence images of the periosteal surface at the proximal, middle, and distal tibial region of those mice at 2 weeks after focal radiation. Endomucin staining (green) indicates blood vessels. NR = nonirradiated tibia; R = irradiated tibia; CB = cortical bone; P = periosteum; M = muscle. Red = Tomato; blue = DAPI. Scale bar = 50 μm. (C) Quantification of Tomato⁺ surface on periosteal surface (Td.S/BS) at three regions. Pro = proximal; mid = middle; dis = distal. (D) Quantification of blood vessel surface on periosteal surface (VS/BS) at three regions. (E) CFU-F assay was performed on the periosteal mesenchymal cells harvested from tibias at 2 weeks after radiation. (F) H&E staining of expanded periosteum (outlined by red lines) at both proximal and distal sides of fracture at 3 days postinjury. Scale bar = 50 μm. (G) Area and thickness of expanded periosteum were quantified. (H) EdU staining of expanded periosteum (outlined by white lines) at both proximal and distal sides of fracture at 3 days postinjury. Scale bar = 50 μm. (I) The percentage of EdU⁺ cells in expanded periosteum was quantified. Values are mean ± SD. n = 6 mice. ^ap < 0.05; ^bp < 0.01; ^cp < 0.001 R versus NR; ^*p < 0.05; ^$p < 0.01; ^#p < 0.001 distal versus proximal. Paired Student’s t test.

The radiation effect on periosteal mesenchymal cells was further confirmed by a 23% decrease in CFU-F colonies derived from a digestion of the entire tibial periosteum in irradiated bones compared with those from nonirradiated bones (Fig. 5E). Note that periosteal cells were collected from the entire bone surface, not only from the irradiated area, likely underrepresenting the effects radiation has on CFU-F ability.Three days after fracture, the periosteal layer in nonirradiated bones was greatly expanded, with greater periosteal area and thickness at the proximal side than at the distal side (Fig. 5F, G). At both sides, irradiated bones displayed significantly reduced periosteal expansion, but the most severe effect was observed at the distal side with 78% and 85% decreases in periosteal area and thickness, respectively, compared with the corresponding site of nonirradiated bones. Consistently, the prior irradiated periosteum had 18% and 83% less EdU⁺ cells at the proximal and distal side, respectively, compared with nonirradiated periosteum (Fig. 5H, I). Taken together, our data demonstrated that prior radiation damages the periosteal mesenchymal progenitors and hampers their proliferation responses toward fracture injury.

Next, cell culture experiments were performed to understand the effects of radiation on periosteal mesenchymal progenitors. In an attempt to mimic the in vivo thickening of the periosteum occurring after fracture injury in the absence of any blood vessels, we cultured cells under either normoxia or hypoxia (0.1% oxygen). As shown in Fig. 6A, and consistent with the above in vivo data, radiation drastically inhibited the proliferation of progenitors. Hypoxia seemed to have no effect on cell proliferation. Interestingly, chondrogenic differentiation ability of these cells was tremendously enhanced by hypoxia condition, resulting in greatly enlarged pellet size (Fig. 6B), more GAG production (Fig. 6C), and highly upregulated chondrogenic marker gene expression (Sox9, Acan, Col2α1, and Col10α1) (Fig. 6E). This might explain why periosteal cells in the region closest to the fracture gap, therefore within the most thickened periosteum after fracture, preferentially differentiate into chondrocytes and periosteal cells in the region away from the fracture gap directly differentiate into osteoblasts. Radiation did not affect hypoxia-induced chondrogenesis. On the contrary, osteogenic differentiation ability of progenitors was greatly reduced by both radiation and hypoxia culture conditions as shown by Alizarin red staining (Fig. 6D), and the expression levels of osteoblastic transcription factors (Runx2 and Osterix) and markers (Ibsp and Osteocalcin) (Fig. 6F). Radiation and hypoxia did not seem to have additive effects. These results suggest that in vivo radiation is unlikely to directly affect the endochondral ossification process during the early stage of fracture repair.

Fig. 6.

The effects of radiation and hypoxia on the proliferation and differentiation of periosteal mesenchymal progenitors in culture. (A) Cell number count shows that radiation suppresses the proliferation of periosteal mesenchymal progenitors. Nor = normoxia; Hyp = hypoxia; NR = nonirradiated; R = irradiated. (B) Images of cell pellets after culturing in chondrogenic medium under normoxia or hypoxia for 2 weeks with or without prior radiation. Scale bar = 1 mm. (C) Alcian blue staining of sections from chondrogenic cell pellets. Scale bar = 50 μm. (D) Alizarin red staining of cells after culturing in osteogenic medium under normoxia or hypoxia for 2 weeks with or without prior radiation. Scale bar = 50 μm(E) Real-time RT-PCR analysis of chondrogenic marker gene expression in cell pellets harvested after 2 weeks of culture in chondrogenic medium. (F) Real-time RT-PCR analysis of osteogenic marker gene expression in cells harvested after 2 weeks of culture in osteogenic medium. Values are mean ± SD. ^ap < 0.05; ^bp < 0.01; ^cp < 0.001 R versus NR; ^$p < 0.01; ^#p < 0.001 hypoxia versus normoxia. Unpaired Student’s t test. All experiments were repeated at least three times.

Fibrotic tissue in fracture nonunion originates from extraskeletal sources

Next, we performed lineage tracing to investigate the source cells for the fibrotic tissue in nonunion fractures. Periosteum is composed of two distinct layers: the outer fibrous layer and inner cambium layer.⁽¹⁹⁾ We observed that Col2-Cre and αSMA-CreER label most cells in the inner layer but is entirely absent in the outer layer. On the contrary, Gli1-CreER labeled a small portion of cells in the inner layer and a majority of cells in the outer layer (Fig. 7A). Among all three Cres, only Col2-Cre marked mesenchymal cells in the endosteum and midshaft bone marrow. At 2 weeks after fracture, in nonirradiated bones, the majority of callus cells, including chondrocytes and osteoblasts, in all three mouse models were Tomato⁺ (Fig. 7Ba, c, e), suggesting that cells within the periosteal inner layer, especially those inner layer cells labeled by Gli1-CreER, are the main contributor for callus formation. Interestingly, in irradiated bones, while the proximal callus was still made of Tomato+ cells, the fibrotic tissue at the distal end was Tomato” in Col2/Tomato and αSMA/Tomato mice (Fig. 7Bb, d), demonstrating that the pathological fibrotic cells did not originate from the cambium layer or from the bone marrow.

Fig. 7.

Fibrotic cells in nonunion fracture originate from extraskeletal tissue. (A) Fluorescence images of tibial sections from 2-month-old Col2/Tomato, αSMA/Tomato, and Gli1/Tomato mice. The latter two mouse lines received daily repetitive tamoxifen injections. The top panel shows the midshaft cortical area and the bottom panel shows bone marrow within the midshaft region. PS = periosteum; ES = endosteum. Red = Tomato (Td); blue = DAPI. Red arrows point to Tomato⁺ cells in the inner layer and yellow arrows point to Tomato⁺ cells in the outer layer of periosteum. Scale bar = 100 μm. (B) Fluorescence images of tibial sections from Col2/Tomato mice, αSMA/Tomato mice at 7 days after fracture, and Gli1/Tomato mice at 14 days after fracture. The latter two mouse lines received daily repetitive tamoxifen injections right before fracture. Dashed yellow lines depict the fracture lines. WB = woven bone; FT = fibrotic tissue; C = cartilage; CB = cortical bone; pro = proximal; dis = distal. Scale bar = 200 μm.

Surprisingly, the fibrotic tissue was Tomato⁺ in Gli1/Tomato mice (Fig. 7Bf). One possible explanation is that Gli1⁺ outer layer cells reconstitute the fibrotic tissue. We believe this is unlikely because we did not observe an expansion of the outer periosteal layer after fracture, regardless of prior radiation. Gli1⁺ cells exist not only in bone but also in many other tissues. Recent studies identified Gli1 as a faithful marker for fibrosis-driving mesenchymal stem cells in solid organs and bone marrow.^(20,21) Therefore, it is likely that the fibrous cells in fracture nonunions comes from non-skeleton-resident Gli1⁺ mesenchymal cells. To test this, we removed a portion of the periosteum at either the proximal or distal side, or both sides of the fracture gap immediately prior to fracture. Two weeks later, normal callus containing cartilage and bones were formed at the site where periosteum was intact but fibrotic tissue was formed at the site where the periosteum was removed (Supplemental Fig. S6).These data provide additional evidence that periosteum is not the source for the fibrotic cells in nonunions.

Discussion

Here, we established and characterized a novel, nonsurgical, traumatic mouse model for atrophic nonunion with high reproducibility. Normal, nonstabilized fractures start with a hematoma quickly formed at the fracture site and intense proliferation of the nearby cambial layer of the periosteum (Fig. 8A). The inner cells of the thickened periosteum close to the facture line then differentiate into chondrocytes, presumably because of the low oxygen tension caused by a lack of vasculature within the thickened periosteum. This process initiates bone repair via the endochondral ossification mechanism. At the region away from the fracture line, cells in the less thickened periosteum preferentially differentiate into osteoblasts to initiate intramembranous ossification repair. However, our data showed that after prior radiation, bone/ cartilage containing callus is only formed at the proximal side of fracture, albeit to a reduced extent, where vascularity is less affected, but not at the distal side (Fig. 8B). Instead, only fibrotic tissue develops at the distal side. The properties of this fibrotic tissue, such as a lack of chondrocyte and bone differentiation abilities and vessel infiltration, are similar to the scar tissue harvested from atrophic fracture nonunion patients. Eventually, the formation of this fibrotic tissue replaces what would otherwise be osseous bridging across the fracture gap and leads to nonunion in our mouse model. Further mechanistic study using a lineage tracing approach demonstrated for the first time, to our knowledge, that mesenchymal progenitors cells from bone tissue, including bone marrow, endosteal, and periosteal mesenchymal progenitors, do not contribute significantly to this fibrotic tissue. Meanwhile, cells from extraskeletal tissues migrate to the fracture site and form fibrosis at the fracture ends that hinders bone repair.

Fig. 8.

A model of how atrophic nonunion fracture is formed after prior radiation. (A) Normal fracture healing starts with hematoma formation at the fracture site and intense proliferation of the nearby periosteum. The inner cells of thickened periosteum close to the facture line then differentiate into chondrocytes and the outer cells remain as periosteal progenitors. At the region away from the fracture line, cells in the less thickened periosteum preferentially differentiate into osteoblasts. Later, calluses formed at both sides of the fracture line merge into one that contains cartilage in the center and bone at the ends. After remodeling, the fracture gap is bridged to restore intact cortical bone. (B) If the fracture site receives prior radiation, the expansion of periosteum shortly after fracture is greatly reduced at the proximal side and almost completely ceased at the distal side, leading to a small callus at the proximal side and fibrosis at the distal site. Fibrous cells are likely derived from extraskeletal tissues. Eventually these two parts cannot consolidate at the fracture gap, resulting in a nonunion fracture.

A remarkably increased incidence of delayed and nonunion fracture within a prior irradiated bone has been a well-known clinical observation for many years. Several previous rodent studies reported that radiation before or immediately after fracture causes alterations in the histology and radiology of the reparative processes.^(22–24) However, all those studies used whole-body radiation, which limited the local radiation dosage that could be applied and was always accompanied by systematic effects, which primarily affect hematopoiesis⁽²⁵⁾ and had abscopal effects on radiation-shielded bones.⁽²⁶⁾ Therefore, they do not faithfully mimic the clinical scenario of the primarily local problem of a traumatic fracture and ensuing nonunion. The focal irradiator used in this study (SARRP) represents a great advantage over the whole-bone irradiator by limiting the radiation field with mm-scale accuracy. Our past experiments have demonstrated that SARRP radiation on one bone has no adverse effects on the contralateral bones.^(10,11) In the current study, we performed bilateral tibial fractures, and the contralateral nonirradiated tibias healed at a similar pace as tibias from mice that have never received SARRP radiation (data not shown). This effectively excludes confounding factors that could result from systemic influences of radiation.

Interestingly, we observed distinct radiation effects on the healing process at two sides of the fracture line in our mouse model. Immediately before fracture, the periosteum in Col2/ Tomato mice seems to be similarly damaged. However, after fracture, the periosteum still thickens, albeit at a much less robust level, at the proximal side, but fails to expand at the distal side. This difference in periosteal response toward injury leads to a smaller, but morphologically normal, callus formed at the proximal site and no callus at all at the distal side, eventually creating an unbridged bony site reminiscent of an oligotrophic process at the proximal side and atrophic one at the distal side. Interestingly, our data are consistent with a previous report that callus formation in rodent tibia is larger at the proximal side than the distal side⁽²⁷⁾ and also consistent with a well-known clinical observation that the rate of nonunion in tibia is location dependent with the highest rate at the distal third part and the lowest rate at the proximal third.^(28–30) In addition, this dichotomous response of our animal model will allow for the concomitant study of the response of both atrophic fibrous and oligotrophic osseous tissues to potential healing/regeneration therapies.

These data demonstrate a pivotal role of the cambial periosteal mesenchymal progenitors in bone repair. Our EdU experiment and lineage tracing in three reporter mouse lines revealed that only cambial periosteal mesenchymal progenitors proliferate and participate in callus formation. Our in vitro experiments showed that radiation mostly affects the proliferation of progenitors but has no effect on their chondrogenic differentiation. Therefore, we conclude that the predominant cellular reason for delayed and nonunion fracture healing is the damage to periosteal mesenchymal progenitors. However, this location-dependent healing in prior irradiated bones indicated that other factor(s) may further affect the periosteal response after injury. Compared with average fracture patients, those with concomitant vascular injury have a drastically increased nonunion rate,⁽³¹⁾ indicating that blood supply is an important contributor for fracture repair. Interestingly, the main artery supplying the periosteal vascular system in the femur and tibia runs from the proximal to the distal end.⁽³²⁾ Our data also showed more periosteal blood vessels at the proximal region of tibia compared with the distal region. Furthermore, at 2 weeks after radiation, vascularization within the irradiated field was still reduced compared with nonirradiated bone. Hence, right after fracture intervention, the distal side should receive much more vascular disruption than the proximal side. After fracture, we have clearly shown a drastic decrease in vasculature around the fracture site in irradiated tibias, particularly at the distal side. We believe that this difference in blood supply is a critical factor that causes the location-dependent injury response of the periosteum. In normal fracture healing, this results in a more thickened periosteum at the proximal site. Under irradiated conditions, this results in greater suppression of the injury response in the distal periosteum. However, eliminating blood supply alone does not cause nonunion. A previous study showed that resection of the femoral artery during fracture in mice still formed bone/cartilage-containing calluses but at a smaller size, producing delayed union.⁽³³⁾

The causes of atrophic nonunion factures in patients are various. Some are known, including prior radiation, large defect size, and periosteal damage, but others are largely unknown. Despite these various causes, a common end feature of this type of nonunion is the fibrotic tissue formation surrounding the fracture ends. This tissue prevents osteocartilaginous tissue growth and is removed during surgical treatment of fracture nonunion. Our mouse model is consistent with this clinic feature, allowing us to better understand the cellular, molecular, and tissue characteristics of regeneration failure, and could aid the development of potential therapies against that failure. Our data strongly suggest that the fibrotic tissue formed in nonunion fracture derives from extraskeletal tissue. First, in Col2/Tomato and αSMA/Tomato mice, the callus, regardless in normal healing or delayed healing, is made of mostly Tomato⁺ cells, whereas fibrotic cells are all negative for Tomato. Second, in the nonunion resulting from periosteum removal, fibrotic cells are only formed at the site where the periosteum is stripped. This is consistent with a previous report that atrophic nonunion forms with fibrous tissue surrounding the fracture site after cauterization of periosteum during fracture in rats.⁽³⁴⁾ Third, we found that Gli1-CreER labels fibrotic cells in the nonunion fracture. Glil is a Zinc-finger transcriptional factor functioning in hedgehog signaling pathway.⁽³⁵⁾ It recently emerges as a marker for mesenchymal stem cells (MSC) located not only in craniofacial⁽³⁶⁾ and appendicular bones⁽⁶⁾ but also in solid organs.⁽²⁰⁾ The fact that it does not label any bone marrow cells in the long bone midshaft, a population of cells that is commonly flushed out for CFU-F assay, suggests that Glil⁺ cells only partially overlap with bone marrow MSCs that are traditionally considered to be important for bone formation. Recently, studies showed that residential Glil⁺ cells are the source of injury-induced fibrosis in multiple organs, including kidney, liver, lung, heart, and bone marrow.^(20,21) Interestingly, we found that the Gli1-CreER-labeled interstitial cells in bone-neighboring muscle are morphologically similar to fibro/adipogenic progenitors (FAPs) cells, another type of residential MSCs responsible for muscle fibrosis⁽³⁷⁾ (data not shown). Considering the close proximity of muscle to cortical bone, we speculate that one possible source of fibrosis in fracture nonunion could be muscle Glil⁺/FAP cells. One limitation of our lineage tracing experiment is that the recombination efficiencies among the three models, one constitutive (Col2-Cre) and the other two inducible (uSMA-CreER and Gli1-CreER), are different. We cannot completely rule out the possibility that Gli1/Tomato mice have a higher Cre recombination efficiency in the periosteum than the other two models, resulting in Tomato signal in the fibrotic tissue after radiation and fracture.

In summary, we established a highly reliable and nonsurgical atrophic nonunion fracture model in mice for future investigations that is relevant to patients undergoing radiotherapy and, more broadly, for those with oligotrophic and atrophic nonunion. Although the specific mechanistic intricacies found in our model have not been fully validated in human disease, we have nonetheless created a foundational system where both fracture (non)healing mechanisms and interventions can be investigated. Consistent with prior clinical observations, our data demonstrate the importance of severe periosteal injury in nonunion formation, suggesting that effective treatment for this deleterious condition should involve restoring or replenishing active periosteal mesenchymal progenitors to the fracture site. In addition, identifying the source of fibrotic tissue, characterizing their properties, and seeking a method to block their generation may pave the way to resolve this clinically challenging disease.