Targeting mitochondrial responses to intra-articular fracture to prevent posttraumatic osteoarthritis

Study design

This study was designed to test whether the two major treatments representing our previous lines of in vitro research, amobarbital and NAC (acute or extended release), could blunt or prevent articular cartilage degeneration after IAF in a porcine hock model of IAF. Power analyses determined that an n = 5 was necessary to provide a greater than 80% chance to detect improvement in automated Mankin scoring based on our previous characterization of the model and associated PTOA (25). Given the expense and duration of these experiments, we decided to combine the NAC and NAC with extended release treatment groups after the first three animals in each group were analyzed because no significant difference was observed between these two NAC treatments. Including the 1-week, 6-month, and 12-month animals, a total of 75 Yucatan minipigs were used for this study. Normal, sham, ORIF, ORIF + amobarbital, and ORIF + NAC groups at the outset contained at least seven animals; however, seven total animals were removed from the study. Data from five animals were removed because of health concerns unrelated to their IAF procedures, and data from two animals were censored when it was found after euthanasia that a small portion of a fracture fixation screw protruded into the joint. In addition, 6 New Zealand White rabbits receiving 2 J impacts to their femoral condyles and 11 New Zealand White rabbits underwent destabilization of the medial meniscus procedures, as previously described (24), and received 10 μM rapamycin 3 weeks after surgery. Criteria for stopping any study before completion included persistent lameness or any indications of discomfort from the animals after surgical recovery not controlled with analgesic treatment. None of the animals met these criteria. All animals were observed by trained personnel on a daily basis and given a score ranging from 0 to 4 during the acute recovery period until the animals were determined to be healthy enough for transport to long-term housing facilities where they received daily monitoring without a pain score. In addition to the animal experimentation, in vitro experimentation includes healthy bovine osteochondral explants harvested from the femoral condyle and impacted with 2 J to assess acute redox metabolic end points, mitochondrial stress responses, and mitochondrial content; healthy bovine chondrocytes extracted from the loaded areas of bovine knees used for assessments of amobarbital’s chemical activity; and human radial head fragments obtained as surgical discards after radial head fractures.

No outliers or other observations were removed on account of their deviation from the mean in any of the experimentation shown. Most end points were chosen at the outset to evaluate the presence of PTOA in the articular cartilage of these animals and thus include numerous histological end points and scoring, as well as several biochemical end points associated with disease. Outcomes related to synovial thickening, subchondral bone thickness, intracellular anabolic function, and oxidative stress were included as the study progressed. Data were collected blindly and decoded only after analysis. Male and female pigs were evenly distributed among groups, although the study was not powered to detect differences between males and females.

Live chondrocyte isolation and tissue culture

For monolayer assays, fresh bovine stifles (knees) obtained intact from a local abattoir (Bud’s Custom Meats) were dissected and digested in collagenase and pronase (0.1 mg/ml; Sigma-Aldrich) overnight. Chondrocytes were then plated after pelleting and resuspension in growth medium. Identical procedures were used for any human chondrocytes harvested from radial head fragments after surgical discard and used in monolayer. Only primary, unpassaged cells were used in this study. For cultures of fresh bovine chondrocytes, as well as cultures of human and bovine osteochondral explants, tissues were maintained using a 50:50 mix of Dulbecco’s modified Eagle’s medium and F12 (Gibco), with 10% fetal bovine serum (Gibco) added. All cultures were maintained at 5% O₂ and 5% CO₂, with the specific exception of the experiment exposing articular tissue to 21% oxygen after impact.

ETC and oxymetric assays

Whole-cell lysates suspended in phosphate buffer were analyzed via a DU800 spectrophotometer (Beckman), as previously described (50). Samples were exposed to 2.5 mM amobarbital (Valeant Pharmaceuticals North America) in culture medium and assayed with increasing concentration of amobarbital present in the assay reactions themselves. Samples assayed without amobarbital present for measurement confirm the reversible inhibition of complex I by amobarbital. All assays conducted in 1 ml total volume. Complex I activity was measured as the rate of rotenone-inhibitable NADH (reduced form of nicotinamide adenine dinucleotide) oxidation. Samples were assayed with or without rotenone (200 μg/ml; Sigma-Aldrich) in working buffer containing 25 mM potassium phosphate buffer, 5 mM magnesium chloride, 2 mM potassium cyanide, bovine serum albumin (BSA) (2.5 mg/ml), 0.13 mM NADH, antimycin A (200 μg/ml), and 7.5 mM coenzyme Q1 (all from Sigma-Aldrich). Complex II samples were incubated with or without succinate in 25 mM potassium phosphate buffer, 5 mM magnesium chloride, 2 mM potassium cyanide, and BSA (2.5 mg/ml) for 10 min at 30°C. After 10 min, antimycin A (200 μg/ml), rotenone (200 μg/ml), 5 mM 2,6-dichloroindophenol (Sigma-Aldrich), and 7.5 mM coenzyme Q1 were added. Complex III samples were assayed in 25 mM potassium phosphate buffer, 5 mM magnesium chloride, 2 mM potassium cyanide, BSA (2.5 mg/ml), 0.5 mM n-dodecyl β-maltoside, rotenone (200 μg/ml), 1.5 mM cytochrome c, and 3.5 mM coenzyme Q2 (all from Sigma-Aldrich). Complex IV samples were assayed in 25 mM potassium phosphate buffer, 0.5 mM n-dodecyl β-maltoside, and 1.5 mM reduced cytochrome c. To evaluate oxygen consumption in the presence of amobarbital by chondrocytes, 20,000 primary bovine chondrocytes per well were plated on XF96 Extracellular Flux Analyzer (Seahorse Bioscience, Agilent) plates; a basal rate was measured, cells were then subjected to different concentrations of amobarbital, and a second rate was measured.

Evaluation of amobarbital toxicity

In accordance with Institutional Review Board regulations, human surgical discards were obtained fresh from patients receiving radial head implants after intra-articular or near-articular fractures of the radial head. These discards are removed from patients at the time of definitive fracture fixation, between 3 and 6 days after fracture, so no treatment effects are expected. Explants were cut from the articular surface of the radial head and cultured for 3 days with or without 2.5 mM amobarbital in normal medium. Half of each explant was homogenized for ATP analyses, and a second half was stained with Calcein AM (Invitrogen) for 30 min and imaged via an Olympus FV1000 confocal laser scanning microscope (Olympus America). ATP analyses were conducted with bioluminescence kits (Sigma-Aldrich) in common usage (51) and normalized to protein concentration, as determined via bicinchoninic acid (BCA) assay or DNA content determined spectrophotometrically.

Porcine IAF model

All animal procedures including rabbit and porcine models were conducted under the approval of the University of Iowa’s Institutional Animal Care and Use Committee (IACUC). Seventy five skeletally mature, 2-year-old Yucatan minipigs of both genders were used. The porcine model (25) uses a severe impact to the bottom (caudal aspect) of the anesthetized animal’s foot to fracture the distal tibia, followed by surgical ORIF of the fracture. A carefully oriented stress-rising saw cut through the anterior tibial cortex ensures a controlled, reproducible IAF when an impaction pendulum drives the talus into the tibia with 40 J [schematic shown in (25)]. This creates a single distal tibia fracture fragment that is then anatomically reduced and fixed using plates and screws during an open surgical procedure identical to human fracture treatment (ORIF). The fractures were anatomically reduced and then plated as previously described (25) using a veterinary-grade, 2.7-mm tibial plateau leveling osteotomy plate (VP4400.R3-2.7, DePuy Synthes). Intra-articular injections of vehicle alone, amobarbital, or NAC for those treatment groups were given after the ORIF procedure was completed, between 30 and 60 min after IAF for each animal depending on pendulum function, as well as individual fixation and suturing durations. To control for the surgical procedure used to create these IAFs, sham surgical animals received the entire procedure with the exception of the fracture-inducing impact; this included internal hardware and a hydrogel vehicle injection without NAC or amobarbital.

For cartilage treatment after IAF, either NAC or amobarbital was dissolved in 0.5 ml of a reverse-thermal responsive hydrogel used for the local delivery (intra-articular injection) of 2.5 mM amobarbital or 10 mM NAC (39, 52, 53). Hydrogel was prepared by adding 17% (w/v) F-127 (Sigma-Aldrich), 0.2% (w/v) hyaluronate (Gel-One, Zimmer Inc.), and 26 mM NaHCO₃ under sterile conditions, tested for endotoxin using the Limulus amebocyte lysate assay (Lonza) before injection. After treatment, animals were casted in a fiberglass cast for 1 week and allowed to bear weight as tolerated. One week postoperatively, casts were removed, a second 0.5-ml injection of the appropriate treatment was given, and casts were replaced for up to three additional weeks.

During this recovery period and in the days preceding euthanasia, animals were given a pain score ranging from 0 to 4. This score was assessed as follows: 0, appetite normal, full weight bearing on IAF limb when standing, full to partial weight bearing on IAF limb when walking, IAF limb touches the floor greater than 75% of the time while under observation; 1, appetite normal, mostly full weight bearing on IAF limb when standing, partial weight bearing on limb when walking, IAF limb touches the floor at least 75% of the time while under observation, interacts with surroundings; 2, appetite normal, partial weight bearing on IAF limb when standing, partial weight bearing on IAF limb when walking, IAF limb touches the floor at least 50% of the time while under observation, interacts with surroundings; 3, appetite normal to slightly decreased, IAF limb is used more for balance when standing/walking (“toe-touches”) than weight bearing, partial weight bearing on IAF limb at least 25% time while under observation, interacts with surroundings; however, will spend more time laying down; 4, decreased appetite, will not stand or has difficulty standing, unwilling to move, nonweight bearing on IAF limb at all times and cannot ambulate around pen. After the final cast removal, animals were left undisturbed in group housing for the duration of the experiments and given normal access to food and water.

Histological preparation and quantitative Mankin scoring

After postmortem tissue harvest, the tibia and talus were split into medial and lateral compartments, fixed in formalin, decalcified, embedded in paraffin, and cut into 5-μm sections. As previously described (25), sagittal sections were analyzed from each joint through the apex of the medial and lateral talar domes, as well as through the deepest concavities on the medial and lateral distal tibia. Each section was costained with Weigert’s hematoxylin and eosin (H&E), safranin O, and Fast Green (all from Sigma-Aldrich) and then digitized on an Olympus VS110 virtual microscopy system (Olympus) through a 20× objective (383 nm per pixel image resolution). This created a whole-joint composite image that was analyzed using a custom-automated Mankin scoring routine (25). This scoring routine assigns a Mankin score for cell density, PG density, and structural defects at every 0.5-mm tissue increment spanning the full articular surface. Score are assigned on the basis of comparison of cell, structural, and image intensity information automatically quantified via image analysis techniques, with reference values defined from normal Yucatan minipig hocks. These scores are then averaged over the full length of the articular surface being analyzed in a manner that approximates modern histological scoring for arthritis [Osteoarthritis Research Society International scores (44, 45)] within the Mankin rubric (46). A standardized 20 mm span of weight-bearing cartilage was evaluated on the talar and tibial surfaces. The automated scoring routines used here also delineated cartilage thickness and the zones of cartilage, allowing quantification of local PG content from safranin O intensities in each zone.

Extended-release NAC preparation and validation

NAC loaded pellets [containing 70% polymer, 30% NAC (Sigma-Aldrich), and 10% plasticizer by weight] were produced using hot-melt extrusion. The particle size of PLGA (50:50; 0.38 dl/g; Absorbable Polymers International) was reduced by grinding in a laboratory-scale grinder until it was the consistency of a fine powder. The PLGA and NAC were mixed in geometric proportions. The mixture was triturated with a pestle to form a uniform mixture for 10 min and subsequently manually granulated with 10% polyethylene glycol (MW 375) for 10 min. The mixture was loaded into a Dynisco extruder hopper, and the pellets were extruded at 90°C at a feeding rate of 200 mg/min, through a ¹/₁₆-inch die circular orifice. After cooling down, the extruded strands of NAC-loaded polymer were cut into pellets weighing 36 mg. The pellets were sterilized by 12-kilogray gamma radiation. The release of NAC from the pellets was measured in phosphate-buffered saline (PBS) by keeping them in an incubator shaker at 37°C and 300 rpm. At each time point, the PBS was removed from the pellet and replaced with fresh PBS. The amount of NAC released at each time point was determined using DTNB [5,5-dithio-bis-(2-nitrobenzoic acid); Sigma-Aldrich], followed by spectrophotometric detection of reduction by NAC at 412 nm.

Oxidative stress measurements

GSH and GSSG were assayed using methods published by Griffith (54). Briefly, samples were minced in 5% sulfosalicylic acid and then subjected to three freeze-thaw cycles to ensure that all cells were lysed. Samples were then added to a buffer containing dithionitrobenzoic acid (DTNB), GSH reductase, and excess NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) (all from Sigma-Aldrich). Reduced GSH molecules react with the DTNB to produce a yellow product, whereupon GSH reductase recycles the GSH. Thus, the rate of absorbance change at 412 nm is proportional to GSH concentration in a relationship determined by standard curve. To measure GSSG, 20% of sample volume of 2-vinylpyridine (Sigma-Aldrich) was added to a sample aliquot for 1 hour before analyses rendering reduced GSH unavailable for recycling.

Nrf2 and p65 IHC was conducted using previously published IHC techniques (39). Briefly, after paraffin embedding and sectioning at 5 μm, samples were stained with primary antibody, anti-Nrf2 (Abcam) or anti-p65 (Abcam), at a dilution of 1:50 for Nrf2 and 1:100 for p65 overnight. In the morning, slides were rinsed with PBS and blocked for 30 min with goat serum and BSA (Gibco), and secondary (goat anti-rabbit) was added (1:250) for 30 min. Secondary was visualized using standard avidin-biotin complex (ABC) and 3,3′-diaminobenzidine (DAB) reagents (Abcam), and slides were scanned using the components detailed for Mankin scoring above. For immunofluorescence (IF), similar protocols to previous studies were used (39); however, visualization was carried out through a goat anti-rabbit secondary conjugated with Alexa Fluor 568 (Thermo Fisher Scientific) and DAPI staining. Slides were imaged via an Olympus FV1000 confocal laser scanning microscope. Quantitation of cell positivity for either IHC or IF was carried out by counting positive cells per 20× objective field and comparing that to the total cell number.

Extracellular flux assays

To determine basal OCR, maximum OCR, and proton leakage rates, standard mitochondrial stress tests were conducted using an XF96 Extracellular Flux Analyzer (Seahorse Bioscience, Agilent), as previously described (11). Injections contained 2 μM oligomycin, 250 nM FCCP, 2 μM rotenone, and 5 μM antimycin A (all from Sigma-Aldrich). OCRs were normalized to cell number, as counted via hemocytometer after trypsinization. Proton leakage rates were normalized to basal OCRs of each individual sample.

Synovial inflammation and monocyte infiltration

The posterior joint capsule was excised and processed into 5-μm-thick axial histological sections and stained with H&E. Slides were digitized on the Olympus VS110 digital histology system, as described above for Mankin scoring. Images from the 1-week survival minipigs were imported into the Visiomorph software (Visiopharm) for analysis of synovial inflammation using an automated routine. Within the user defined area, cell nuclei were automatically identified within the adipose tissue based on color thresholds, and monocytes were separated from adipocytes based on nuclear aspect ratio. Monocyte count was converted to monocyte density by dividing the cell number by the area of tissue analyzed. In the synovial tissue harvested 6 and 12 months after IAF, the adipose tissue was fully fibrosed, and synovial thickening had to be measured manually. For these specimens, measurement tools included in the software used to run the digital image acquisition (VS-ASW, Olympus) were used to manually measure the thickness of the synovial tissue at five locations distributed over the full tissue section. These five replicate measurements were averaged to obtain a single thickness measurement for each section. Subchondral bone thickness measurements were made manually using the software-based measurement tools at anterior, middle, and posterior locations on the same sagittal histological images used to derive the Mankin scores.

Live-cell analysis of mitochondrial content

Assessment of bovine osteochondral explants for total mitochondrial content after impact was carried out using 200 nM MitoTracker Deep Red (Invitrogen) for 30 min. Live-cell staining and imaging (costained with Calcein AM) were identical to the human staining above. Images within an area impacted with 2 J via 1-kg drop tower (14) were captured via an Olympus confocal microscope using a 4× or 10× objective, and ImageJ was used to quantify average intensity in the damaged area. Treatments applied individually to culture medium after impact only included 2.5 mM amobarbital and 10 mM NAC. For experiments using 200 μM α-tocopherol, α-tocopherol acetate (Sigma-Aldrich) was used to facilitate entry into cells; explants were treated 24 hours before impact and 24 hours after impact. Treatments preloaded 1 hour before impact included 200 μM trolox, 2.5 mM dimethylmalonate, and 2.5 mM dimethylsuccinate (all from Sigma-Aldrich).

Statistical analyses

For all experimentation shown, error bars indicate SDs. All statistical analyses were completed using GraphPad Prism 7. P < 0.05 was used to determine significance, except where otherwise noted. Column comparisons shown in Figs. 1 and 5 indicate one-way ANOVA. Comparisons in Fig. 4 indicate two-sided Student’s t test between the noted groups and control, and pain score differences at 6 months shown in fig. S5 were compared using a χ² analysis. Differences noted in Figs. 2, 3, and 7 were determined with two-way ANOVA including treatments and normal and sham animals, where appropriate, followed by posttesting using Dunnett’s test to compare the effects of multiple treatments. Monocyte infiltration data and proportions of cells staining positive for p65 from Fig. 6 were tested using two-sided Student’s t tests between hydrogel and NAC/glycyrrhizin-treated groups and untreated groups, and no significant differences were found.