Materials and Methods

Animals—Twenty-four horses were used in the study. All horses were 2 or 3 years old. This group of horses was used in another study¹² conducted to test treatment with shock waves and PSGAG concentration. At the onset of the present study, none of the horses had evidence of musculoskeletal carpal disease as determined via radiographic and physical examinations. All procedures were approved by the Colorado State University Animal Care and Use Committee.

Experimental procedures—Horses were allowed to acclimate to the facility and treadmill exercise for 14 days prior to induction of osteoarthritis. The day of surgery for induction of osteoarthritis was designated as day 0. Arthroscopy was performed on both the left and right middle carpal joints of each horse, and the joints were confirmed to be healthy with no visible pathological changes. One middle carpal joint in each horse was randomly selected for osteoarthritis induction via creation of an osteochondral chip as described elsewhere.⁵ Briefly, an 8-mm osteochondral fragment was created in the distal aspect of the radiocarpal bone. A burr was used to remove 15 mm of cartilage from the area of fragmentation; the debris was left in the joint. Incisions for the arthroscopic portals were sutured, and bandages were placed on both carpi.

Treatment and exercise—Horses were allocated into 3 groups for postsurgical treatment. One group (control group) was treated via application of shock waves to both middle carpal joints. Joints initially received 2,000 pulses, which subsequently was reduced to 1,500 pulses, with the probe head covered in air cushioning material^a to block the pulses. A second group (shock wave–treated group) received the same shock wave treatment but without the probe head covered, which allowed the pulses to penetrate the tissues. The third group (PSGAG-treated group) received injections of PSGAG (500 mg, IM, q 4 d for 7 injections). Results for these horses have been reported elsewhere.¹² In that study, statistical analysis of the peripheral WBC data for these 3 groups revealed no significant (all values were P > 0.05) differences in gene expression among the control and 2 treatment groups. Therefore, these data were combined into 1 group for microarray analysis in the present study.

After arthroscopy, the horses were allowed stall rest for 13 days. On day 14, they began treadmill exercise, which consisted of trotting (16 to 19 km/h) for 2 minutes, followed by galloping (32 km/h) for 2 minutes, and then trotting again for 2 minutes. Horses completed the exercise 5 d/wk until day 70 of the study. At that time, the horses were euthanized by IV administration of an overdose of pentobarbital.

Sample collection—Blood and synovial fluid samples were collected for analysis before surgery to induce osteoarthritis (day 0) and on days 14, 42, and 70 after surgery. An additional blood sample was collected on day 7 and used only for microarray analysis.^b Blood samples (2 mL/sample) were collected from the left jugular vein and placed into blood tubes specialized for RNA processing^c and into anticoagulant-free tubes for analysis of protein markers. The tubes were centrifuged to separate blood components, and then the serum and WBC buffy coat layer were harvested separately and stored at −80°C until used for protein and RNA analysis, respectively. Synovial fluid was harvested aseptically from the middle carpal joint of the osteoarthritic limb and stored at −80°C until used for analysis.

Immediately after the horses were euthanized on day 70, cartilage was harvested in a sterile manner from the osteoarthritic joint of each horse. Cartilage samples were collected from the radiocarpal bone adjacent to the area of the osteochondral defect, the opposing surface of the third carpal bone, and an unrelated area on the ulnar carpal bone. Synovial membrane samples were collected from the dorsal aspect of the middle carpal joint. These tissue samples were frozen at −80°C in a lysis reagent^d until used for RNA isolation and analysis of gene expression.

Serum and synovial fluid biomarkers—Protein biomarker assays were performed with the serum and synovial fluid samples obtained from the 8 control horses on days 0, 14, 42, and 70. Collagen synthesis was measured with an ELISA kit^e against CPII. Type II collagen degradation was measured with an ELISA against the equine-specific collagenase degradation product, Col CEQ. These ELISAs were validated for use in equine samples.^13–15

RNA isolation—The RNA from all 24 horses was isolated with an extraction solution^f; RNA was isolated from cartilage and synovium samples collected from the radial, ulnar, and third carpal bone. Briefly, the frozen cartilage was pulverized with a cryogenic mill.^g The extraction solution was added to the cartilage and synovium samples, and they were then homogenized on ice. The RNA extraction protocol was used to isolate RNA, which was then purified with a commercially available kit.^h The RNA was harvested from WBCs as described in the protocol for a blood RNA kit.^c The purity and concentration of the RNA was assessed via a spectrophotometer.ⁱ Some aliquots of WBC RNA were used for the microarray analysis, and the rest were reverse transcribed^j into cDNA for use in the PCR assay. A bioanalyzer^k was used to further determine the purity and concentration of the RNA for microarray analysis.

Microarray analysis and PCR assay—An equine-specific custom microarray^b that contained 3,100 unique equine gene sequences was used for analysis. The RNA from peripheral blood WBCs was used to estimate mRNA expression in samples obtained from all horses on days 0, 7, 14, 42, and 70.

On the basis of the microarray data, PCR primers and probes were created for candidate genes that were significantly differentially expressed in peripheral WBCs when results for samples obtained on days 7, 14, 42, and 70 were compared with results for day 0 (Appendix). Samples for the 8 control horses collected from various cartilage locations in the osteoarthritic limb (ie, radial, ulnar, and third carpal bone), synovial membrane, and peripheral WBCs was used to estimate the mRNA concentrations via real-time PCR assay. Sample cDNA and standardized DNA for each gene were combined with the respective primers and probes, Taq polymerase, nucleotides, and buffer. The cDNA and standards were then amplified and quantified in a thermal cycler. Results for glyceraldehyde 3-phosphate dehydrogenase were used to normalize the data, and CT^−ΔΔCT values were used to determine the change from baseline (the value on day 0 for WBCs and the highest value on day 70 for each tissue group). The microarray data from WBCs were compared with PCR data for the WBCs and with protein biomarker concentrations at days 0, 14, 42, and 70. The PCR data for the cartilage and synovial membrane samples at day 70 were compared with the protein biomarker concentrations.

Statistical analysis—All data were tested for a normal distribution via a Studentized residuals plot and the Shapiro-Wilk test with a software program.^l The microarray and protein biomarker data were logarithmically transformed to normalize the data because the Studentized residuals plot appeared skewed or the Shapiro-Wilk score was < 0.05. All other data met the conditions for a normal distribution without transformation. Data points were eliminated if they were > 2.5 SDs from the mean value. A maximum of 1 data point was removed from each data group.

Microarray values for the 24 horses were compared over time via an ANOVA with the Holm step-down procedure to control the familywise error rate for the 3,100 genes on each chip. A Bonferroni adjustment was performed to identify days on which values differed significantly. The individual expression on days 0, 7, 14, 42, and 70 was used for this analysis.

Microarray data were compared with PCR values via linear regression analysis. Only data for the 8 horses in the control group were used in this analysis; data were available for days 0, 14, 42, and 70. Biomarker values for the serum and synovial fluid samples were compared with the PCR values for each of the tissues via regression analysis of day 70 values for each horse. Again, only data for the 8 control horses were used in this analysis. In a similar manner, biomarker values for serum and synovial fluid samples were compared with the microarray gene profiles via regression analysis. Significance was defined as values of P ≤ 0.05 (2-sided analysis) for all tests.

Results

Microarray analysis—Ten of the 3,100 genes evaluated in WBCs were significantly differentially expressed over time in horses with osteoarthritis (ADAMDEC1 [P = 0.001], GRP94 ([P < 0.001], HCST [P < 0.001], hUNC-93A [P = 0.05], RRM2 [P < 0.001], BM735567.V1.3 at [P < 0.001], WBC008E06 V1.3 at [P = 0.002], BM781012.V1.3 at [P = 0.005], WBC597. gRSP.V1.3s at [P < 0.001], and WBC018F02 V1.3 at [P < 0.001]). Five of these genes were also differentially expressed in horses with laminitis, recurrent airway obstruction, and equine herpesvirus infection.^m The 5 genes differentially expressed only in horses with osteoarthritis were ADAMDEC1, GRP94, HCST, hUNC-93A, and RRM2. Gene expression and days on which values for these 5 genes differed significantly were determined (Figure 1).

Mean ± SD microarray expression of genes that encode for ADAMDEC1 (A), GRP94 (B), HCST (C), hUNC-93 (D), and RRM2 (E) in peripheral WBCs of horses with experimentally induced osteoarthritis. Day of surgery to induce osteoarthritis was designated as day 0. Values on the y-axis represent mRNA expression. Notice that the scale for the y-axis differs among panels. ^a–cFor each gene, values with different letters differ significantly (P ≤ 0.05).

Citation: American Journal of Veterinary Research 74, 1; 10.2460/ajvr.74.1.115

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Mean ± SD microarray expression of genes that encode for ADAMDEC1 (A), GRP94 (B), HCST (C), hUNC-93 (D), and RRM2 (E) in peripheral WBCs of horses with experimentally induced osteoarthritis. Day of surgery to induce osteoarthritis was designated as day 0. Values on the y-axis represent mRNA expression. Notice that the scale for the y-axis differs among panels. ^a–cFor each gene, values with different letters differ significantly (P ≤ 0.05).

Citation: American Journal of Veterinary Research 74, 1; 10.2460/ajvr.74.1.115

• Download Figure
• Download figure as PowerPoint slide

Mean ± SD microarray expression of genes that encode for ADAMDEC1 (A), GRP94 (B), HCST (C), hUNC-93 (D), and RRM2 (E) in peripheral WBCs of horses with experimentally induced osteoarthritis. Day of surgery to induce osteoarthritis was designated as day 0. Values on the y-axis represent mRNA expression. Notice that the scale for the y-axis differs among panels. ^a–cFor each gene, values with different letters differ significantly (P ≤ 0.05).

Citation: American Journal of Veterinary Research 74, 1; 10.2460/ajvr.74.1.115

• Download Figure
• Download figure as PowerPoint slide

Comparison of mRNA from microarray analysis of peripheral WBCs versus protein biomarker concentrations in serum and synovial fluid—The WBC gene expression data from the microarray analysis were compared with Col CEQ and CPII concentrations in serum and synovial fluid samples. Data for days 0, 14, 42, and 70 were used for the regression analysis. Correlations of the 5 genes used for PCR assay were included (Table 1). Serum concentrations of Col CEQ were significantly correlated with expression of GRP94 (R² = 0.174; P = 0.047) and RRM2 (R² = 0.280; P = 0.001) over time.

Comparison of mRNA from PCR assay of cartilage and synovium versus protein biomarker concentrations in serum and synovial fluid—Expression of the 5 genes of interest in cartilage and synovium samples obtained from osteoarthritic joints on day 70 was compared with Col CEQ and CPII concentrations in serum and synovial fluid samples obtained on day 70. Significant correlations were detected for gene expression in the articular cartilage and concentrations of CPII in synovial fluid and concentrations of Col CEQ in serum. Changes in gene expression in the synovium were correlated with CPII concentrations in synovial fluid (Table 2).

Comparison of peripheral WBC mRNA via microarray analysis and PCR assay—To validate use of the microarray for the 5 genes of interest, gene expression for the microarray analysis was compared with gene expression for the PCR assay in WBCs obtained on days 0, 14, 42, and 70. There was not a significant correlation in results between the microarray analysis and PCR assay for HCST (R² = 0.428; P = 0.118) and GRP94 (R² = 0.228; P = 0.424). However, there was a significant correlation in results between the microarray analysis and PCR assay for ADAMDEC1 (R² = 0.652; P = 0.002), hUNC-93A (R² = 0.693; P = 0.003), and RRM2 (R² = 0.549; P = 0.007).

Discussion

The study reported here was designed to identify via use of a microarray the genes that were differentially expressed (gene signature) in horses with osteoarthritis. Analysis of the data indicated 10 genes were differentially expressed; however, only 5 genes were unique to osteoarthritis because the others were differentially expressed in other diseases of horses.^m Because only 3 other diseases were evaluated (laminitis, recurrent airway obstruction, and equine herpesvirus infection), it is possible that these genes could also be increased in other diseases of horses.

Thus, we attempted to validate the expression of ADAMDEC1, GRP94, HCST, hUNC-93A, and RRM2 as measured with microarray analysis and PCR assay. We were unable to establish a significant correlation for GRP94 and HCST. The 3 likely causes for this discrepancy were an inaccurate microarray value, an inaccurate PCR value, or the limited number of horses. The microarray involved use of an algorithm that compared perfectly matched probes with 1-base–mismatched probes to determine the specific amount of RNA for each gene. This algorithm has been faulted for having a high degree of variance because it does not normalize the overall probe intensity between samples.¹⁶ Alternatively, the PCR assay was validated in our laboratory through primer and probe testing by use of standards made of plasmids in which the desired sequence was inserted. Because these primers and probes had not been used previously, we were not familiar with their behavior with native equine cDNA. The microarray data were for 24 horses in which only 8 (control) horses were used for PCR and protein biomarker assays. This was to ensure that there was no bias with confounding treatment of the 16 other horses and, in part, because of economic constraints for the study. The difference in horses may have contributed to the lack of significant findings.

When the differentially expressed genes (gene signature) associated with osteoarthritis were investigated further, there were significant decreases in expression of 3 of the 5 genes after the onset of osteoarthritis (Figure 1). Evaluation of upregulation or downregulation of these genes was beyond the scope and design of the present study. However, gene expression was compared with concentrations of biomarkers in an attempt to determine whether there was a relationship.

We detected variations in the correlation of gene expression and concentrations of protein biomarkers that was dependent on the tissue examined. Overall, there was a poor correlation of the protein biomarker concentrations with gene expression in WBCs over time. There was a significant decrease in gene expression for most of the genes after the onset of osteoarthritis, whereas concentrations of all of the protein biomarkers increased significantly after the induction of osteoarthritis.⁵ The reason for correlations between only gene expression and serum concentrations of Col CEQ and synovial fluid concentrations of CPII, but not between gene expression and synovial fluid concentrations of Col CEQ and serum concentrations of CPII, is unknown. One likely cause could have been the difference in sensitivities of the ELISAs used for measurement of protein biomarker concentrations. Another cause for the lack of a correlation could be that protein biomarker concentrations are globally independent of gene expression patterns. Future examination of the relationship between the gene signature and concentrations of putative biomarkers may be warranted.

The link between the expression of some genes and osteoarthritis was more intuitively logical than that for other genes. This was likely attributable to incomplete knowledge of the systemic effects of osteoarthritis and the intercellular signaling cascade that allows for disease progression.

The ADAMDEC1 is a protein in the subgroup of the ADAM gene family, which includes the aggrecanases.¹⁷ Aggrecanases are pivotal in the enzymatic cleavage of aggrecan, which is one of the most important extracellular matrix components of cartilage because of its ability to withstand compressive forces.¹⁸ Aggrecanases are increased in osteoarthritic cartilage of humans¹⁸ and horses.¹⁹ The ADAMDEC1 is released from macrophages and dendritic cells and plays a role in activation of the immune system.²⁰ Macrophages are important cell mediators in the progression of osteoarthritis.¹⁸

Through an increase in GRP94, there is a subsequent increase in the amount of immunoglobulin released from monocytes and a dose-dependent increase in monocyte proliferation.²¹ This reflects an increase in the innate immune response and may cause increased numbers of macrophages to be available for mediation of osteoarthritis in diseased joints.²¹

Hematopoietic cell signal transducer is a transmembrane protein that acts in the activation of T cells and natural killer cells.²² Natural killer cells and T cells are present in higher numbers in the synovium of humans with osteoarthritis than in normal synovium, and those cells potentiate inflammation associated with osteoarthritis.²³

The Unc-93 is a transmembrane protein that may form an ion channel, and hUNC-93A is a homolog of Unc-93.²⁴ To our knowledge, there has been no information published relating the behavior of Unc-93 to osteoarthritis.

Ribonucleotide reductase M2 polypeptide is a subunit of ribonuclease reductase that controls the activity of nuclear factor-κB, a transcription factor that serves to regulate many cell activities.²⁵ Included in these activities is regulation of matrix metalloproteinase-9, a metalloproteinase that plays a part in extracellular breakdown of cartilage during osteoarthritis.²⁶

One of the benefits of the present study was that a large number of candidate genes were screened in peripheral WBCs, which we believe could represent systemic changes when there is local disease in joints. The corresponding problem is that this type of analysis can fail to identify a gene that is upregulated locally but not systemically. Nevertheless, we were able to identify 5 genes that were significantly differentially expressed in the peripheral WBCs, even though only a solitary joint in each horse was involved.

It should be recognized that although collection of blood samples is easy to perform and a relatively non-invasive procedure, use of a microarray is complex, requires specialized equipment, and is currently expensive. One of our goals was to identify a subset of genes that could be screened with a less expensive, more clinically applicable system. These gene expression patterns may need to be coupled with protein patterns to determine clinical relevance.

Gene expression of ADAMDEC1, GRP94, HCST, hUNC-93A, and RRM2 in peripheral WBCs of horses were useful as gene biomarkers for experimentally induced osteoarthritis because their expression was altered by the induction of osteoarthritis. Future studies should be conducted to test the gene expression profile function in horses with naturally developing osteoarthritis. Although the osteochondral fragment technique has been commonly used to induce osteoarthritis in horses, clinically affected animals are likely to have some variation as a result of the duration and severity of disease. Eventually, the gene expression profile could be used as a diagnostic aid in screening to identify horses with osteoarthritis. It may become an easy method for use in differentiating joint disease from other causes of lameness.

In the present study, 5 genes were identified that were differentially expressed in horses with experimentally induced osteoarthritis. Three of the 5 genes identified via microarray assay were confirmed via PCR assay. Repeat analysis to confirm the other 2 genes is necessary before their clinical application can be evaluated. Expression of the 3 genes (ADAMDEC1, hUNC-93A, and RRM2) should be confirmed in horses clinically affected with osteoarthritis to investigate their use in clinical applications.