All procedures were approved by the Institutional Animal Care and Use Committee. The mice used had a targeted disruption of the Thbs2
gene, which encodes the thrombospondin-2 protein (TSP2-null).(24
) Coisogenic WT 129/SvJ mice were used for comparison.
We created closed, transverse fractures in both tibias of 63- to 70-day-old mice using methods similar to those described previously by Hiltunen et al.(25
) Briefly, mice were anesthetized for all surgical procedures using isoflurane (Aerrane; Baxter, Deerfield, IL, USA), and 0.05 mg/kg of butorphanol tartrate (Torbugesic-SA; Fort Dodge Animal Health, Fort Dodge, IA, USA) analgesic was administered subcutaneously shortly after anesthetic induction. Both legs were prepared for aseptic surgery. Mice were placed in dorsal recumbency on microwaveable heating pads for the duration of anesthesia to maintain normal body temperature. The stifle joint of the right leg was flexed, and a small incision was made just medial to the tibial tuberosity. A 26-gauge hypodermic needle was used to bore a hole in the cortex of the medial aspect of the tibial tuberosity, slightly distal to the stifle joint. A sterile, 0.009-in-diameter, stainless steel pin was inserted into the created hole and inserted down the length of the tibia in the intramedullary canal until resistance was felt, indicating full insertion. This served as an intramedullary pin that would provide stability at the fracture site. This procedure was repeated for the left leg. After pin insertion, the pins were cut to be flush with the cortex, and the skin defect was closed using tissue adhesive (Nexaband; Abbott Laboratories).
Fractures were created in both legs using a custom-made device that uses a sliding weight and guillotine mechanism. This device produces consistent controlled displacement, high-energy impact force sufficient to induce fractures in mouse tibias. Mice were placed in sternal recumbency, and each leg was individually placed in the guillotine and fractured. Whole body radiographs were generated using a microradiography system (Faxitron, Wheeling, IL, USA) to verify pin placement and fracture gap location. Fractures analyzed were typically midshaft, simple, transverse fractures, although occasionally fracture occurred in the distal one third of the tibia. Tape “splints” were placed on both tibias to provide initial rotational stability to the fracture region for the first 48 h.
Mice recovered after the procedure under heat lamps. Moistened food was placed on the cage bottom, and water was provided ad libitum. Mice were typically ambulatory within 1 h after surgery and were observed eating within a few hours. Mice were maintained in a cage with wireless tops to reduce climbing activity. No mortality was observed throughout the course of this study.
Tissue harvest and preparation
At harvest, all animals were anesthetized with isoflurane gas anesthetic and humanely killed by cervical dislocation. Right tibias were carefully dissected, the intramedullary pins were removed, and the tibias were placed in 4% paraformaldehyde for 24 h, decalcified in formic acid for 12 h, and transferred to 70% ethanol until further processing for histology or immunohistochemistry (IHC). Left tibias were similarly dissected, wrapped in saline-soaked gauze, and placed in storage at −20°C until μCT scanning and torsional mechanical testing could be performed.
Samples were scanned using an eXplore Locus SP microCT system (GE Healthcare Preclinical Imaging, London, Ontario, Canada) and reconstructed at an 18-μm isotropic voxel size using the Feldkamp cone beam algorithm. A custom software analysis procedure was specifically developed to quantify the callus properties on these images using Microview (v 2.1.2 Advanced Bone Application; GE Healthcare Preclinical Imaging) similar to that described by Den Boer et al.(26
) First, the image was reoriented so that the anterior-posterior and longitudinal axes were aligned with the principal image axes. In the second step, three independent reviewers scrolled through the image planes and measured the maximum callus width using a line that bisected the middle of the marrow cavity as well as the maximum callus length on the anterior side of the bone (). The measurements for maximum callus width and maximum callus length were averaged across the three reviewers, and the average length was used to isolate the callus from the image of the entire bone (). Next, the callus and cortical bone sections were manually segmented using a series of user-defined points with spline interpolation between these points (). The points for cortical bone boundary were chosen on slices of the image not more than 30 CT slices (0.540 mm) apart, and spline interpolation was used to define the points in between. These points were reviewed and modified, and a reinterpolation was performed in an iterative process. A similar process was used to define the callus boundary. Next, a single point within the cortical region of interest was used to initiate a region-growing algorithm that detected the cortical bone by finding all connected voxels over a simple global threshold. This region-growing algorithm was confined by the cortical region of interest to ensure that mature bone within the callus, particularly near the proximal and distal ends, was not included in the cortical bone measurements (). The cortical bone voxels were removed from the image so that it did not bias any measurements (). Last, the region of interest surrounding the callus was identified (), a global threshold was applied, and the callus volume, bone volume fraction, BMD, BMC, tissue mineral content (TMC), and tissue mineral density were calculated (TMD). Bone mineral measurements represent the mineral contained in the entire callus volume. Tissue mineral measurements represent the mineral contained within the volume defined as bone ().
FIG. 1 Quantification of callus length, total volume, and bone volume using μCT. (A) Anterior callus length was measured in all specimens (white line). (B) Next, using the maximum callus length, the callus was cropped from the image. (C) Splines were (more ...)
Tibias were secured in brass pots using a low melting temperature Cerro Alloy (McMaster Carr, Chicago, IL, USA) and mounted into a custom torsion testing device. This torsion tester was equipped with a 50 in.oz. reaction torque sensor (Model 2105–50; Eaton, Troy, MI, USA) and an RVDT (Model R30A; Lucas Control Systems, Hampton, VA, USA) for torque and angular displacement measurements, respectively. Raw torque data were conditioned with a strain gage amplifier (2100; Measurements Group, Raleigh, NC, USA), and angular displacement was conditioned with an LVDT amplifier (DTR-451; Lucas Control Systems) before collection. This device was interfaced with LabVIEW (v 7.0; National Instruments, Austin, TX, USA) for data collection and controlled using a custom program that interfaced using a data acquisition system (NI PCI-6251; National Instruments). The bones were tested at a constant displacement rate of 0.5°/s until failure while being maintained moist at room temperature. Data were sampled at 1000 Hz at a displacement rate of 0.5°/s and stored for analysis. Analysis was performed using a custom MATLAB (v 7.0.1; The Mathworks, Natick, MA, USA) script. In this script, the torque data were filtered with a third-order Savitzky-Golay FIR smoothing filter with a 0.5-s window before analysis to remove noise. The stiffness was calculated based on a linear regression on the torque-displacement data in a user-selected region, and the script automated calculations for torque at failure, angular displacement at failure, and energy to failure.
Safranin-O staining was performed on right mouse tibias that were paraffin embedded and serially sectioned (7 μm). Briefly, slides were deparaffinized, rehydrated, and exposed to 0.3% Fast Green FCF (Fisher, Pittsburgh, PA, USA). Slides were rinsed in 1% acetic acid (Fisher), immersed in 5.45% Safranin-O (Fisher), rinsed in dH2O, dehydrated, mounted using Permount (Biomeda, Foster City, CA, USA), and visualized using a microscope. These samples were used for measuring total callus area, chondrocyte area (i.e., Safranin-O–positive area), and area of woven bone. Hypertrophic chondrocyte areas were defined based on the characteristic appearance of those chondrocytes.
Right mouse tibias were paraffin embedded and serially sectioned (7 μm). Sections were next deparaffinized, rehydrated, immersed in heated citrate buffer, and incubated with 3% H202 in PBS. Slides were blocked and exposed to the respective primary antibody: collagen type IIa (graciously gifted from Dr. Linda Sandell, Washington University), proliferating cell nuclear antigen (PCNA; Chemicon International, Temecula, CA, USA), and von Willebrand's factor (vWF; Dako, Carpinteria, CA, USA). After treatment with the primary antibody (collagen type IIa, 1:5000; PCNA, 1:1500; vWF, 1:400), sections were treated with biotin-conjugated secondary antibody (for collagen type IIa, 1:150; for PCNA, 1:400; for vWF, 1:500; for PCNA: Jackson Immunoresearch, West Grove, PA, USA; for collagen type IIa/vWF: Vector Laboratories, Burlingame, CA, USA). After secondary antibody treatment, slides were treated with streptavidin-conjugated horseradish peroxidase (HRP; StriAviGen Super Sensitive Label Antibody; Biogenex Laboratories, San Ramon, CA, USA) and diaminobenzidine (Dako). Slides were counterstained with Gill's hematoxylin, dehydrated, mounted using Permount (Biomeda, Foster City, CA, USA), and visualized using a microscope. Mouse spleen and heart tissue were used as control tissues for vWF, and mouse testicular tissue was used as positive and negative control tissue for PCNA.
Expression of β-galactosidase, osteocalcin (OCN), osterix (OSX), Sox9, TSP2, and VEGFA was assessed by immunofluorescence in serially sectioned (7 μm) paraffin-embedded fractured tibias samples (β-galactosidase/OSX/Sox9; Abcam, Cambridge, MA, USA; OCN: Takara, Shiga, Japan; TSP2: BD Transduction Laboratories, San Jose, CA, USA; VEGFA: Novus, Littleton, CO, USA). After treatment with the primary antibody (β-galactosidase, 1:250; OCN/OSX/Sox9/TSP2/VEGFA, 1:100), sections were treated with Alexafluor 594–labeled secondary antibodies (1:200; Molecular Probes, Carlsbad, CA, USA), mounted with Vectashield containing DAPI (Vector Laboratories, Burlingame, CA, USA), and visualized with a fluorescent microscope.
After tissue rehydration, slides were placed in 95°C citrate buffer for 20 min and, after removal, allowed to cool for 20 min. Next, slides were incubated for 15 min in 3% H2O2 and incubated for 2 h with the terminal transferase (TdT; Roche, Basel, Switzerland) and biotin-conjugated 16-dUTP (Roche) reaction mixture. Incubation with streptavidin-conjugated HRP and DAB chromogen and hematoxylin counterstain were carried out in the same manner as was performed for IHC. Mouse testicular tissue was used for controls. Positive control tissues were incubated with DNase. Negative controls used testicular tissue without the TdT and or without the 16-dUTP. All control tissues were run in parallel with samples.
TSP2 and control β-galactosidase adenovirus were generated by the University of Michigan Vector Core. TSP2-null mice tibias were fractured, and at day 3 after fracture, 10 μl of 1 × 108 TSP2 adenovirus or LacZ control adenovirus particles was injected into the fracture site using Luer Tip Hamilton syringes. Each mouse was injected with LacZ on one side and TSP2 on the contralateral side. The mice were given 10 days to heal after fracture and killed. Tissue was collected and processed as described in the IHC methods and stained according to the Safranin-O protocol detailed earlier.
Total callus area, Safranin-O–positive area, hypertrophic cell area, woven bone area, and the area of cells showing collagen type IIa collagen expression were measured at ×4 magnification using Bioquant Image Analysis software (Bioquant Image Analysis, Nashville, TN, USA). Using the manual measure function of this software, we identified appropriate areas and outlined them. Three tissue sections per slide were measured, and the average was calculated.
For the quantification of PCNA and TUNEL labeling, we used a method similar to that described by Li et al.(2
) Briefly, the length of the fracture callus was measured using the Bioquant software at ×4 magnification. Within the proximal, middle, and distal third of the fracture callus, the total number of positive and negative cells was measured in three fields of view at ×63 magnification for a total of nine fields per tissue section. These measurements were repeated on three tissue sections on the same slide for a total of 27 measures per sample. The average number of positive cells over all fields measured represents the percentage of positive cells within the fracture callus.
Total OCN, OSX, Sox9, and VEGFA areas were quantified, at ×200 magnification, using the SigmaScan Pro 5 software (Aspire Software International, Ashburn, VA, USA). Using the manual measure function of this software, positive areas within the total callus were identified and outlined. Three tissue sections per slide were measured, and the average was calculated.
Time course gene expression analysis for TSP2
WT (C57/B6) mice were placed in groups (n
= 4 per time point) and given either 0 (no fracture), 5, 7, 10, 14, 18, or 21 days to heal and were subsequently killed. Fracture calluses were carefully dissected and immediately snap frozen in a liquid nitrogen bath. Frozen tissue samples were homogenized using a liquid nitrogen-cooled mortar and pestle apparatus, and mRNA was purified using TRIzol (Invitrogen, Carlsbad, CA, USA). Single-strand cDNA was synthesized from exactly 0.5 μg mRNA from each sample. Specific primers for β-actin (internal control) and TSP2 were used for real-time PCR analysis (Corbett Research, Carlsbad, CA, USA). Samples were denatured at 94°C for 20 s, annealed at 58°C for 30 s, and amplified at 72°C for 30 s for a total of 30 cycles. C(t) results were compared with β-actin expression and fold-change (relative to day 9 nonfracture controls) was determined using previously described methodology.(22
QPCR gene expression analysis for day 5 samples
Fracture calluses were extracted and macerated using a Tissue Tearor (Biospec Products, Bartlesville, OK, USA). RNA was isolated using the Qiagen RNeasy Mini kit (Qiagen, Valencia, CA, USA). RNA yield was determined spectrophotometrically, and integrity was confirmed by gel electrophoresis. cDNA synthesis was done as described previously. The expression levels of the following genes were determined using a 7500 Fast Real-Time System machine (Applied Biosystems, Foster City, CA, USA): collagen type IIa, OCN, OSX, Runx2, Sox9, VEGFA, β-actin, and laminA. Taqman Gene Expression Assays (Applied Biosystems) were used for OSX, Runx2, Sox9, VEGFA, and LaminA (internal control), and the following primers were used for collagen type IIa, OCN, and β-actin (internal control) with Power SYBR Green (Applied Biosystems) [Forward (F)/Reverse(R)]:
- collagen type IIa: (F)GGCTCCCAGAACATCACCTA/(R)TCGGCCCTCATCTCTACATC
- osteocalcin: (F)CGCTCTGTCTCTCTGACCTC/(R)TCACAAGCAGGGTTAAGCTC
- β-actin: (F)AAGAGCTATGAGCTGCCTGA/(R)TGGCATAGAGGTCTTTACGG
RNA fold expression levels were calculated using the double ΔCT method, and proper amplicon formation was confirmed by melt curve analysis.
One-way ANOVA was used to assess statistical significance between WT and TSP2-null samples at each time point. Results for the temporal analysis of gene expression were analyzed using one-way ANOVA with Tukey posthoc analysis. Animal numbers for each experiment varied and are indicated in the figure legends.