Reactive oxygen and nitrogen species induce protein and DNA modifications driving arthrofibrosis following total knee arthroplasty

The clinical records of patients were reviewed in detail to extract variables including age, sex, body mass index (BMI), years post initial surgery, pre-existent co-morbidities, functional scores (particularly the details of range of motion) and all other relevant information (Table 1). Ten patients undergoing uneventful primary TKA were matched for age, sex and BMI and included as the control cohort.

There was no correlation between most of the demographic factors, range of motion and the molecular findings in this cohort of patients. The only significant correlation observed was between range of motion (ROM) and tissue calcification/bone volume (BV) (Table 1) [13]. As no other correlations were evident, all subsequent analyses were grouped based on patient tissue BV. The control tissue BV was 0.005 ± 0.01.

Histological analysis of the 10 control tissues showed ~1% average inflammatory cell presence (Figure 1A). This is in contrast to the arthrofibrotic tissues, which showed a 3.0 - 4.0-fold increase in the number of macrophages and a 3.0 - 9-fold increase in lymphocytes as compared to control (Figure 1B, C, D). The inflammatory response did not include the infiltration of neutrophils (Figure 2). The absence of these cells suggests that the inflammation was not due to infection. Image analysis of the fibrotic regions (non-calcified areas) showed increased inflammation in patient tissues in both the low and high BV groups (Figure 1D). The increase in macrophage and lymphocyte numbers were significant in the both BV groups relative to control tissue (P < 0.05).

Based on the increased presence of macrophages in the arthrofibrotic tissues, subsequent analyses to determine the levels of MPO were performed. MPO is an enzyme that produces highly reactive products that mediate chlorination, protein/DNA hydroxylation and protein/DNA nitrosylation [36‐41]. Low levels of MPO were detected in control tissue macrophages by immunohistochemistry. A representative tissue image is shown in Figure 3A. Figures 3B and 3C show the elevated expression of this enzyme in macrophages within the arthrofibrotic tissue. Although less intense, MPO expression was also observed in the periarticular fibroblasts, which was not observed in control tissues (Figure 3D). Image analysis revealed a correlative increase in macrophage (based on CD68 image analysis) and fibroblast expression of MPO in arthrofibrotic patient tissues (Figure 3E). This increase was statistically significant for the high BV group compared to the control group (P < 0.05) and the low BV group approached significance.

Based on the increased MPO in the arthrofibrotic tissues, subsequent analysis were performed in order to determine the levels of ROS/RNS. As direct measurement of ROS/RNS production in tissues lacks sensitivity and reproducibility, we measured protein nitrosylation and DNA hydroxylation (8-OHdG), both are end products of ROS/RNS-mediated reactions. There was no detectable protein nitrosylation (Figure 4A) or 8-OHdG modifications (Figure 5A) in control tissues by immunohistochemistry. Evident within the fibrotic regions of the arthrofibrotic tissues were nitrosylated proteins (Figure 4B and 4C) and cells containing 8-OHdG (Figure. 5B and 5C). The ROS/RNS-mediated 8-OHdG nuclear modification was localized to regions of high fibroblast density and was significantly increased in tissues in the high BV group (Figure 5D; P < 0.05).

In order to determine MPO expression specific to the arthrofibrotic fibroblast population and to analyse oxidative stress, we performed a real-time polymerase chain reaction (PCR) microarray panel for oxidative stress responsive genes (Table 2). In agreement with the immunohistochemical finding, MPO expression levels in the isolated arthrofibrotic fibroblasts were 2.4-fold higher than control fibroblasts. In contrast to MPO, the expression of the major anti-oxidant enzyme superoxide dismutase 1 (SOD1) was down-regulated 2.9-fold. A number of other oxidative stress responsive genes were also disregulated, showing a significantly increased or decreased expression by arthrofibrotic fibroblasts. Of note was a 7.6-fold increase in polynucleotide kinase 3'-phosphatase (PNKP), an enzyme involved in the base excision repair pathway for nearly all ROS/RNS-induced DNA mutations, and an 11.9-fold increase in thioredoxin reductase 1 (TrxR1) expression, an oxidoreductase enzyme that promotes cell proliferation. Finally, there was 21.1-fold decrease in the expression of another important anti-oxidant, microsomal glutathione S-transferase 3 (MGST3). SOD1 is a both a cytosolic and a secreted protein, and MGST3 is a major intracellular antioxidant enzyme.

This multi-centre study used a standardized tissue retrieval protocol allowing collection and analysis of periarticular tissues from the knee of patients undergoing revision arthroplasty for arthrofibrosis. The diagnosis of arthrofibrosis is based on clinical, radiological examination and intra-operative findings [11]. For these patients, the distinct intra-operative findings are extensive fibrotic tissue formation that fills the lateral, medial, and parapatellar gutters, generally within 1 year after TKA. In this study, tissue samples from 10 affected knees and 10 knees from osteoarthritis (OA) patients undergoing primary TKA (controls) were retrieved. Primary surgical tissues controls were chosen for comparison with arthrofibrotic tissues, as these tissues represent the pre-surgical status, where normal OA associated inflammation exists but is not associated with excessive fibrosis or metaplastic changes [65].

Tissue samples were taken from the periarticular area, which included the suprapatellar, medial gutter, lateral gutter and infrapatellar regions. The tissue was wrapped in sterile saline soaked gauze and transferred, or shipped overnight on ice, to the laboratory for fixation and detailed analyses. Tissue from each anatomical location was cut into 2 × 5 mm pieces and, depending on the amount of available tissue, four to five pieces of tissue from one region were placed in a paraffin block. An equal number and distribution of tissue cubes was used for fibroblast isolation. Any remaining tissue was flash frozen in liquid nitrogen and stored at -80°C. Tissue collection was performed in accordance with the Institutional Review Board guidelines of the participating institutes.

Each of the paraffin blocks containing tissue were subjected to μCT analysis in order to determine heterotopic ossification (Scanco μCT 40, Basserdorf, Switzerland), with an energy of 45 kVp, a current of 88 μA and a 200-ms integration time producing a resolution of 20 μm³ voxel size. Each scan comprised a minimum of 500 slices through the entire paraffin block. In order to achieve image noise reduction, a constrained three-dimensional Gaussian filter (sigma 1.2, support 2) was applied. A fixed, global threshold for analysis was chosen that represented the transition in X-ray attenuation between un-mineralized tissue (< 225) and the forming bone (230 - 700). Analysis consisted of defining the outer boundary of the tissue for each 20 mm section in the sample. For consistency, the same settings and thresholds were used for each analysis and applied to every sample in the study. Scout, sagittal and cross-sectional views were examined for evidence of mineralization.

Tissues were fixed in 4% paraformaldehyde, dehydrated, embedded in paraffin and sectioned (6 μm). Paraffin sections were dewaxed, rehydrated and stained with Harris Hemotoxylin (Fisher Scientific, MI, USA; Nos 245-678) and Eosin Y (Fisher Scientific; Nos 245-827) in order to determine cellularity, vascularization and tissue morphology. Sections were also stained with Wright Geimsa (Fisher Scientific; Nos 264-985; phosphate buffer pH 6.8 Nos 262-237) in order to determine the inflammatory cell number and toluidine blue (Sigma, MO, USA; No. 198161) to determine mast cell numbers.

Paraffin sections (3 μm) were mounted on Fisher Superfrost/Plus slides which were placed in a 58°C oven for 30 min prior to immunostaining. A Ventana Benchmark XT automated slide stainer (Ventana, AZ, USA; N750-BMKXT-FS) was used for the following immunohistochemical staining reactions. The slide stainer was equipped with an iView DAB detection kit (Ventana) for immunoperoxidase visualization of the targeted antigen. Endogenous biotin reactivity in the tissue sections was blocked using the Endogenous Biotin Blocking kit from Ventana. After the completion of the staining run, the slides were briefly washed in a mild dishwashing detergent solution (Dawn; Proctor & Gable, Ohio, USA) to remove the liquid cover slip solution and processed for haematoxylin counterstaining using a 1:8 dilution of Gills-3 haematoxylin solution (Polysciences, PA, USA) for 1 min. Slides immunostained for 8-OHdG antigen were counterstained with Fast Green Substitute for Light Green Working Solution (Poly Scientific, NY, USA). Specific staining conditions were as follows: p53 mouse monoclonal anti-human antibody (clone: Bp53-11; pre-diluted, 37°C for 32 min) (Ventana; 760-2542) (CC1 treatment* - 1 h); CD68 mouse monoclonal anti-human antibody (clone: KP-1; pre-diluted, 25°C for 32 min) (Ventana; 790-2931) (CC1 treatment* - 30 min); neutrophil elastase mouse monoclonal anti-human antibody (clone: NP57; 1:100 dilution, 25°C for 1 h) (Dakocytomation; M-0752) (no antigen retrieval); 8-OHdG monoclonal antibody (1:20 dilution, 25°C for 1 h) (Oxis International, Inc, CA, USA; 24328) (CC2 treatment* - 8 min then Protease 3 treatment - 16 min); myeloperoxidase mouse anti-human antibody (1:300 dilution, 25°C for 1 h) (Dakocytomation, CA, USA; A-0398) (CC2 treatment* - 1 h); and rabbit anti-nitrotyrosine (1:1000 dilution, 37°C for 1 h) (Harry Ischiropoulos, University of Pennsylvania, USA) (CC2 treatment* - 1 h);. *CC2 is a citrate based buffer, pH 6.0, Protease 3 is a low concentration solution of a serine protease and CC1 is a Tris buffer with 1 mM EDTA (antigen retrieval reagents, Ventana; 950-124). Slides were then mounted in Permount, cover slipped and evaluated by microscopy. Control tissues to determine antibody reactivity and conditions included, human tonsil tissue, bone marrow, normal breast tissue and breast tumour tissue.

For each patient two to three blocks of tissue from each anatomical site were sectioned and complete images of each section (25-30 individual images) were acquired at a magnification of 20×. Images were acquired with a Retiga EXi digital-cooled camera with a red, green and blue electronic filter (QImaging, BC, Canada) or with an RT Color Spot camera (Diagnostic Instruments, MI, USA) on either a Nikon Optiphot or a Nikon E800 (Nikon, NY, USA). Image quantification was performed with Image Pro Plus software (Mediacybernetics, MD, USA), using a customized macro to count diaminobenzidine (DAB) stained cells and nuclei of cells stained with haematoxylin. A quantitative value of the inflammatory response was then presented as the average percent of positive cells (DAB) per total cell number (haematoxylin) normalized to total area. The section results for each block from each anatomical site were averaged and site differences compared.

In order to evaluate the number of macrophages (CD68) and MPO positive cells, the number of each in serial sections was compared. If they existed, the counts were corrected for total cell number and total area of the section.

Equal amounts of tissue were taken from each region and combined (total weight ~0.5 g). The 2 × 2 cm pieces were rinsed in Hanks' balanced salt solution to remove any blood, placed in a 25 ml flask and incubated at 37°C for 4 days in α-MEM with 10% fetal bovine serum, 100 units/ml streptomycin and 100 units/ml penicillin (Mediatech, Inc, VA, USA) [66]. Tissue pieces were removed and the media changed on day 4. Cells were passaged when they reached approximately 80% confluency. Cells were not used beyond passage seven.

RNA isolation, quantitative SYBR^® Green-based real-time PCR (RT-PCR) oxidative stress and antioxidant defense array and data analysis were performed by SABiosciences (Frederick, MD, USA). Fibroblasts isolated from arthrofibrotic tissue and from primary knee tissue (control) were analysed. The expression level of the three housekeeping genes, β2-microglobulin, β-actin and glyceraldehyde-3-phosphate dehydrogenase, were used to normalize the data presented.

Statistical analysis between groups was performed using a one way ANOVA for normality and student's t-test for continuous variables. A level of significance (α), or a P-value of less than 0.05, with a 95% confidence interval, was determined. In order to evaluate correlations for patient clinical information and individual test values for each patient, a Pearson correlation coefficient was calculated. For pairs with P-values of less than 0.05, there was a significant relationship between the two variables. All parameters were evaluated with SPSS software (version-Base 13.0; SPSS, IL, USA).