Rush University Medical Center and the Hospital for Special Surgery have filed a provisional US patent application based upon results presented in this article.
Traumatic and degenerative meniscal tears have different anatomic features and different proposed etiologies, yet both are associated with the development or progression of osteoarthritis (OA). In established OA, synovitis is associated with pain and progression, but a relationship between synovitis and symptoms in isolated meniscal disease has not been reported. Accordingly, we sought to characterize synovial pathology in patients with traumatic meniscal injuries and determine the relationships between inflammation, meniscal and cartilage pathology, and symptoms.
Thirty-three patients without evidence of OA who were undergoing arthroscopic meniscectomy for meniscal injuries were recruited. Pain and function were assessed preoperatively; meniscal and cartilage abnormalities were documented at the time of surgery. Inflammation in synovial biopsy specimens was scored, and associations between inflammation and clinical outcomes were determined. Microarray analysis of synovial tissue was performed, and gene expression patterns in patients with and those without inflammation were compared.
Synovial inflammation was present in 43% of the patients and was associated with worse preoperative pain and function scores, independent of age, sex, or cartilage pathology. Microarray analysis and real-time polymerase chain reaction revealed a chemokine signature in synovial biopsy specimens with increased inflammation scores.
Our findings indicate that in patients with traumatic meniscal injury undergoing arthroscopic meniscectomy without radiographic evidence of OA, synovial inflammation occurs frequently and is associated with increased pain and dysfunction. Synovia with increased inflammation scores exhibit a unique chemokine signature. Chemokines may contribute to the development of synovial inflammation in patients with meniscal pathology; they also represent potential therapeutic targets for reducing inflammatory symptoms.
Joint injury predisposes individuals to develop osteoarthritis (OA) (1, 2). Among the most common knee joint injuries associated with increased OA risk are meniscal injuries. Recent longitudinal data from the Multicenter OA Study indicate that meniscal damage is associated with a 6-fold increased risk (odds ratio 5.7, 95% confidence interval [95% CI] 3.4–9.4) of developing radiographically visible OA changes (3). Furthermore, in patients with established OA, meniscal damage is associated with risk of progression (4). Anatomic patterns of meniscal tear are often used to discriminate between traumatic and degenerative meniscal pathology; traumatic tears occurring in an otherwise normal meniscus are reported to present with longitudinal (sometimes “bucket-handle” type tears) or radial orientations, while horizontal, flap, or complex tears and maceration are interpreted as degenerative tears, i.e., those occurring in a meniscus structurally weakened by degenerative change (5). Both patterns of meniscal alteration are associated with an elevated risk of OA (6–8), but the risk associated with degenerative-type tears appears to be higher (9). Biomechanical factors play a role in the structural changes in both patterns of meniscal pathology, but cellular and molecular processes that lead to an increased risk of OA are not understood. Furthermore, these injuries are often asymptomatic (10), and factors contributing to symptoms such as pain have not been defined.
In patients with OA, inflammation is one factor associated with risk of both progression of cartilage loss (11, 12) and symptoms (13–15). Inflammation in OA joints manifests as synovial membrane (SM) mononuclear cell infiltration observed in early and late stages of disease (16–20). However, it is unclear whether inflammation predates or is a consequence of OA development. Roemer and colleagues (21) recently noted an association between meniscal damage and synovial effusion on magnetic resonance imaging (MRI), but the cellular and molecular nature of this inflammation was not defined. Pessler et al (22) noted a mild synovitis with histologic features similar to OA in a group of patients with “orthopedic arthropathies,” including some with meniscal tears. However, prevalence of inflammation in patients with meniscal injuries in the absence of preexistent OA has not been well studied.
The present study was designed to define the prevalence and characteristics of synovial inflammation in patients undergoing arthroscopic partial meniscectomy for traumatic meniscal injury in the absence of antecedent evidence of OA. Furthermore, we sought to determine whether synovial inflammation is associated with preoperative clinical symptoms. A histologic scoring system to grade inflammation was validated using independent evaluators, and comparisons were made with previously characterized synovial tissue from patients with OA.
PATIENTS AND METHODS
The study was approved by the Institutional Review Board of the New England Baptist Hospital (NEBH), and all patients provided written informed consent. Patients ages 18–60 years who had a traumatic knee injury and were scheduled for arthroscopic partial meniscectomy for treatment of symptomatic meniscal tears were recruited from the NEBH Department of Orthopedic Surgery. The inclusion criteria were patient recall of a knee injury that initiated the symptoms and occurred within 6 months of presentation, and a meniscal tear identified on preoperative MRI and considered to be the cause of symptoms. We excluded those with known inflammatory arthritis or symptoms to suggest systemic inflammatory arthritis (i.e., >30 minutes of morning stiffness, multiple joint complaints, concurrent back pain), patients with radiographic evidence of OA (osteophytes or joint space narrowing), and patients with meniscal tears affecting the vascular portion of the meniscus thought to be amenable to surgical repair rather than resection. The latter was done to increase the homogeneity of the patient population. For comparative evaluation of histopathology, meniscectomy patients were compared with a group of 20 patients with knee OA whose synovium had been previously biopsied (20). These patients included 14 patients with advanced knee OA who met American College of Rheumatology (clinical and radiographic) criteria (23) and had undergone total knee replacement surgery and 6 patients with earlier-stage knee OA who had undergone arthroscopic surgery with intraoperative evidence of cartilage loss or fibrillation, but without advanced radiographic changes (Kellgren/Lawrence radiography scores of ≤2) (24).
The Short Form 12 (SF-12) health survey (25) and Lysholm questionnaires were administered preoperatively. In addition, patients were asked to assess their knee pain on a visual analog scale (VAS). The Lysholm questionnaire is a knee-specific instrument for measuring symptoms (pain, swelling, limp, locking, instability) and functional disability (stair-climbing, squatting, use of supports) on a single scale (0–100). Originally developed to assess responses to ligamentous repairs (26), this instrument has been validated in patients undergoing meniscal procedures (27, 28). In contrast, the SF-12 captures information on general physical and emotional well being.
Assessment of meniscus and cartilage integrity.
Surgical reports were available for 28 patients, and were reviewed to determine the anatomic pattern of meniscal pathology (degenerative versus traumatic). Cartilage integrity was assessed intraoperatively by the operating surgeon and noted in operative reports using the Outerbridge scoring system (29), where 0 = normal articular cartilage, 1 = superficial softening, 2 = superficial fissuring or fibrillation involving <1.25 cm area, 3 = fibrillation or fissuring involving >1.25 cm area, and 4 = full-thickness cartilage wear with exposed subchondral bone. When operative reports were not available (n = 5) or cartilage condition was not recorded clearly (n = 6), videos taken during surgery were reviewed by the principal surgeon (BM), who determined Outerbridge scores. The worst score from all 3 compartments (medial, lateral, and patellofemoral) was used in this analysis.
Synovial tissue collection and preparation.
Tissue from patients undergoing meniscectomy was obtained from 3 locations: the suprapatellar pouch and medial and lateral gutters. Biopsy specimens were obtained from areas of synovium that appeared to be inflamed or thickened. When no inflammation was apparent, biopsy specimens were obtained from standard locations: the femoral aspects of the medial and lateral gutters, and the central supratrochlear region in the suprapatellar pouch. Biopsy specimens from patients with knee OA had been obtained from the suprapatellar pouch only (20). Tissue biopsy specimens were formalin fixed and paraffin embedded before sectioning and hematoxylin and eosin (H&E) staining.
Histologic assessment of synovial inflammation.
Both meniscectomy and OA synovial specimens were subjected to the same assessment protocol. To standardize evaluations, we analyzed only sections containing a clearly recognizable lining layer with underlying vascularized subintima, and evaluated inflammation at low power (5× objective). Since there are no published reports on synovial infiltrates in patients with meniscal injury only, inflammation was graded based on perivascular mononuclear cell infiltration in synovium from OA patients (20, 30), where 0 = none, 1 = mild (0–1 perivascular aggregates per low-power field), 2 = moderate (>1 perivascular aggregate per low-power field with or without focal interstitial infiltration), and 3 = marked aggregates (both perivascular and interstitial). To evaluate interreader and intrareader reliability, 18 synovial specimens were scored by 2 independent readers (ED and CRS) and 8 were rescored by 1 reader (ED) under blinded conditions.
Synovial gene expression microarray analysis.
Eight synovial biopsy specimens were chosen for microarray analysis, 4 each from meniscectomy patients with synovial inflammation (grade 1 or 2) and without synovial inflammation (grade 0). The biopsy specimens were from different patients, and anatomic locations varied. Total RNA was extracted from homogenized SM samples using PerfectPure RNA Fibrous Tissue kits (5Prime Inc.). All RNA was DNase treated, oligo(dT) primed, and complementary DNA (cDNA) was synthesized with Superscript III reverse transcriptase (Invitrogen Life Technologies). RNA integrity was determined by electrophoresis on a microfluidics-based platform (Agilent Technologies). RNA was hybridized to Affymetrix human U133 plus 2.0 chips at the Weill Cornell Medical College Core Facility (New York, NY). Data were analyzed using Genespring 10.0 software (Agilent Technologies) as follows. Raw data were transformed using the robust multichip analysis algorithm with baseline transformation to the median of all arrays. Probe sets were filtered by expression (20–100%), with the requirement that probes be present in at least 4 of 8 arrays. An unpaired t-test was performed on the filtered data. We found 3,030 probe sets differentially expressed in synovial inflammation samples (P < 0.05); 260 were differentially expressed with a >2-fold difference. Pathway overrepresentation analysis was performed using algorithms available via the Innate DB database (http://www.innatedb.ca/index.jsp) (31). Innate DB is a database of genes, proteins, interactions, and signaling responses involved in mammalian innate immune responses. Targets were then chosen for validation by real-time quantitative polymerase chain reaction (PCR).
Quantitative PCR analysis.
After accounting for the histologic analysis, there was sufficient tissue available from 37 biopsy specimens for analysis of gene expression by real-time PCR. Twelve of the 37 specimens were suprapatellar, 14 were from the medial and 11 from the lateral gutter, and the specimens were obtained from 18 patients. RNA was extracted and cDNA synthesized as described for the microarray analysis. Messenger RNA (mRNA) levels of 4 chemokines and 1 chemokine receptor identified by pathway analysis of the microarray data (interleukin-8 [IL-8], CCR7, CCL19, CCL21, and CCL5) were measured by real-time PCR using specific primers (sequences available from the author upon request) and iQ SYBR Green Supermix (Bio-Rad). Primers spanned introns and yielded a single product. After normalizing Ct values to GAPDH, expression levels were calculated relative to the mean of specimens without inflammation.
Interreader and intrareader reliability of inflammation scores is reported as a weighted kappa statistic. Given the small sample size and some irregularly distributed variables, nonparametric tests were used. Between-group differences were evaluated with Mann-Whitney t-tests, and Spearman's correlation coefficients were calculated using Prism 5.0 software (GraphPad Software). Generalized estimating equations (GEEs) with an unstructured correlation matrix were applied when multiple data points per patient were analyzed (relative expression levels), to adjust for lack of independence of the data. Multiple linear regression analysis was performed to examine the association between suprapatellar inflammation score and baseline Lysholm scores. Age, sex, body mass index (BMI), and time between injury and surgery were included as independent covariates.
Thirty-three patients who met the criteria were recruited. Demographic characteristics of these patients and the OA patients used for comparison are presented in Table 1. For meniscectomy patients, the median interval between knee injury and surgery was 14.8 weeks (range 1–42 weeks). Most (79%; n = 26) had medial meniscal tears; 6 had lateral tears, and 1 had both medial and lateral tears. Surgical reports were available for 28 patients; 23 indicated the presence of complex tears (with horizontal cleavages and flap lesions, 1 described as macerated). Only 1 had a single radial tear, and 2 had both medial radial and lateral complex tears, while in 2 reports the type of tear was not recorded. Despite exclusion of patients with radiographic OA, only 7 patients (21%) exhibited Outerbridge grade 0 (normal) cartilage in all compartments. The remainder had grade 1 (n = 6), grade 2 (n = 7), or grade 3 (n = 7) lesions in ≥1 compartment, with 6 exhibiting focal grade 4 chondral lesions but no diffuse full-thickness cartilage loss. Median BMI was similar in meniscectomy and OA patients, but OA patients were older (median age 64 years in OA patients versus 45 years in meniscectomy patients; P < 0.0001) and more likely to be female (P < 0.05 by Fisher's exact test).
Table 1. Characteristics of the patients undergoing partial meniscectomy and the patients with OA*
Partial meniscectomy patients (n = 33)
OA patients (n = 20)
OA = osteoarthritis; BMI = body mass index; NA = not applicable.
P < 0.0001 versus meniscectomy patients, by unpaired t-test.
Patients with Kellgren/Lawrence scores of >0 were excluded from the meniscectomy group.
Based on assessment of suprapatellar biopsy specimens.
Findings of histologic evaluation of synovial inflammation.
In 5 cases, biopsy specimens were mishandled or mislabeled during specimen procurement and preparation, resulting in only 28 patients contributing biopsy specimens sufficient for evaluation. Inflammation was graded on a scale of 0–3 in both meniscectomy and OA patients. Interreader and intrareader weighted kappa statistics for the inflammation score were 0.87 and 1.0, respectively. Figure 1 shows photomicrographs of biopsy specimens from representative meniscectomy patients with typical grade 0, 1, and 2 inflammation. None exhibited grade 3 inflammation.
Prevalence and anatomic variation of inflammatory infiltrates.
The 3 locations biopsied in meniscectomy patients were compared first. Inflammation was observed most often in suprapatellar biopsy specimens (12 of 28 [43%]), compared with medial or lateral gutters (7 of 27 [26%]). When suprapatellar inflammation was observed, it was often found in ≥1 gutter as well (7 of 12). Five patients exhibited suprapatellar inflammation only; 2 had inflammation only in the gutters. When results were analyzed according to side (medial or lateral) of meniscal injury (ipsilateral or contralateral) there was no predilection for inflammation in the gutter on the side of the meniscal pathology (data not shown). We next compared the extent (grade) and prevalence of synovial inflammation in meniscectomy and OA patients. Since biopsy specimens from OA patients had been obtained from the suprapatellar pouch (30), comparison was made only at this location. Inflammation was observed less often in meniscectomy patients than in OA patients (43% versus 75%) (Table 1), and tended to be of lower grade.
Association of inflammation with patient characteristics and Lysholm scores.
Meniscectomy patients were stratified according to presence (grade 1–2; n = 12) or absence (grade 0; n = 16) of suprapatellar inflammation. Lower Lysholm scores (indicating greater knee-related symptoms/disability) were observed in patients with suprapatellar inflammation than in patients without (difference between means −19.9 [95% CI −9.20, −30.7], P = 0.0008). No significant differences in SF-12 scores (−0.85 [95% CI 1.08, −2.79]) or VAS scores (0.44 [95% CI 2.27, −1.40]) were observed. Patients with inflammation were significantly older (mean ± SD 51.3 ± 7.3 years versus 40.2 ± 11.6 years; P = 0.007), and the interval between injury and surgery was significantly shorter for patients with inflammation (mean ± SD 10.2 ± 8.8 weeks versus 18.5 ± 11.5 weeks; P = 0.047). Inflammatory infiltrates were observed in some patients presenting for surgery within a few weeks of their reported injury.
Although the difference in mean Outerbridge cartilage scores was not significant, patients with synovial inflammation tended to have higher Outerbridge scores (2.3 ± 1.2 versus 1.3 ± 1.5; P = 0.07). Only 1 of 7 patients with normal (grade 0) cartilage demonstrated inflammation. Of the 6 patients with focal grade 4 lesions, 5 were female, but otherwise these 6 were not clearly distinguishable from the rest, and Lysholm scores varied widely (range 40–90). Synovial biopsy specimens were available for 4 of these 6 patients; 2 exhibited synovial infiltrates (grade 1), and 2 did not. There was no correlation between Outerbridge scores and Lysholm scores (r = 0.03, P = 0.86).
Multiple linear regression analysis was performed to determine whether the relationship between suprapatellar synovial inflammation and Lysholm scores was independent of known OA risk factors and degree of cartilage abnormality. Age, sex, Outerbridge score, BMI, and time between injury and surgery were included as independent covariates. Both inflammation score (P = 0.001, effect estimate −15.3 ± 4.7 per point) and BMI (P = 0.004, effect estimate −1.3 ± 0.4 per kg/m2) were significantly associated with Lysholm score after adjusting for the above variables. Outerbridge score (P = 0.69) and age (P = 0.30) were not.
Analysis of synovial gene expression in patients with and those without synovial inflammation.
Figure 1 shows H&E-stained sections from one biopsy specimen without inflammation (Figure 1A) and one biopsy specimen with inflammation (Figure 1C) subjected to microarray analysis of gene expression patterns. Two hundred sixty genes were differentially expressed by ≥2-fold (P < 0.05) between biopsy specimens with and those without inflammation. Inflammatory pathway overrepresentation analysis (31) of differentially expressed genes revealed 22 “pathways” (transcripts that cluster into functional categories or molecular pathways) that were significantly enriched, with corrected P values of less than 0.05. We focused on the 7 clusters that included >3 gene products (Table 2). Of these 7, a signature of chemokines and their receptors was the most highly up-regulated pathway in biopsy specimens exhibiting inflammation. The 6 transcripts in this signature are shown in Table 3 with their respective fold change and P values.
Table 2. Pathway analysis of the 260 transcripts differentially expressed in synovial biopsy specimens with and those without inflammation*
Differentially expressed genes
Pathway names and P values were obtained from the InnateDB database (http://www.innatedb.ca/index.jsp). Differentially expressed genes were determined by microarray analysis.
In biopsy specimens with inflammation versus those without inflammation.
Comparison of biopsy specimens with inflammation and those without inflammation, by unpaired 2-tailed t-test.
Of the other pathways identified (Table 2), “primary immunodeficiency” and “hematopoietic cell lineage” were composed of cell surface receptors and genes associated with infiltrating leukocyte populations (i.e., CD19, IL2RG, IL7R, CIITA, CD1D, and CD2). Three additional pathways included overlapping lists of cytokine receptor chains (IL2RB and IL2RG), an intracellular signaling molecule (JAK3), and cytolytic enzymes (GZMA and GZMB) expressed by T and natural killer cell populations. This was expected since we had defined inflammation as perivascular mononuclear cell aggregates, which are largely composed of lymphocytes (30). The seventh pathway, “cytokine–cytokine receptor interactions,” comprised the same 6 chemokine/receptor transcripts and cytokine receptor chains (IL2RB and IL2RG). For the purpose of the present analysis, we focused our attention on the chemokines because of their potential contribution to early events in lymphocyte accumulation in synovium.
Validation of chemokine expression by real-time PCR.
Levels of mRNA for 4 chemokines and 1 chemokine receptor identified by microarray pathway analysis (IL-8, CCR7, CCL19, CCL21, and CCL5) were measured by real-time PCR. All available biopsy specimens yielding sufficient cDNA quantities were used (37 samples representing 18 patients). Samples were stratified by inflammation score (absence or presence of inflammation), and relative analyte expression levels were compared. IL-8 (Figure 2A), CCL5 (Figure 2B), CCR7 (Figure 2C), and CCL19 (Figure 2D) were all detected more frequently in biopsy specimens exhibiting inflammation. Each patient contributed up to 3 biopsy specimens in this analysis, so GEE models were run on log-transformed data to adjust for lack of independence. The GEE model revealed statistically significant relationships between inflammation and CCL19 (P = 0.0096) and CCL5 (P = 0.0307), and trends toward a significant association with IL-8 (P = 0.0649) and CCR7 (P = 0.066). CCL21 was undetectable in most specimens (data not shown).
Association of chemokine expression with baseline Lysholm scores.
Associations between chemokine expression in suprapatellar biopsy specimens and clinical outcome scores were assessed by Spearman's correlation. CCR7 expression (Figure 3A) and CCL19 (Figure 3B) expression showed strong negative associations with Lysholm scores (r = −0.790, P = 0.002 and r = −0.867, P = 0.002, respectively) in meniscectomy patients. Higher Lysholm scores indicate a less symptomatic knee. IL-8 (r = −0.54, P = 0.07) and CCL5 (r = −0.38, P = 0.2) were moderately but not significantly associated with Lysholm scores. No associations between chemokine expression and VAS pain or SF-12 scores were observed.
Emerging evidence indicates that synovitis is related to OA symptoms and progression (11–15). Synovial inflammation and effusions also occur with meniscal injuries (21), even in patients without radiographic OA. However, cellular and molecular characteristics of synovitis associated with meniscal damage have not been reported. We sought to determine the prevalence and molecular features of synovial inflammation in patients who did not have preexistent radiographic features of OA and were undergoing arthroscopic meniscectomy for clinically documented traumatic knee injury with MRI evidence of meniscal tears. Specifically, we wanted to determine whether synovial inflammation correlated with symptoms.
A previous study (22) demonstrated similar synovial pathology in OA patients and patients with joint injury. We therefore compared histologic features of synovial inflammation in meniscectomy patients to those in patients with established knee OA (30). Appearance of cellular infiltrates was similar, but inflammation was less prevalent and extensive in meniscectomy patients. Contrary to what we anticipated, we did not see preferential localization of inflammation in the gutter on the side of the meniscal tear. One possible explanation is that we did not analyze the perimeniscal SM directly adjacent to the tear. However, of the 3 locations biopsied, inflammation was identified most often in the suprapatellar location (in 43% of patients), suggesting that synovial inflammation occurs within the joint at sites distant from the injury and is not only localized adjacent to the injury. Our findings are consistent with the results of a recent study of the anatomic distribution of synovitis in knee OA defined with contrast-enhanced MRI techniques (32). In that study, the suprapatellar region was the second most common area (59.5% of patients) in which synovitis was detected. The explanation for involvement of the suprapatellar region remains unclear. We speculate that certain sites within the joint may be uniquely sensitive to the effects of proinflammatory factors produced in response to meniscal injury.
We investigated whether inflammation was associated with preoperative joint symptoms and dysfunction. When stratified according to the presence or absence of suprapatellar inflammation, Lysholm scores were significantly lower (P < 0.05) in patients with inflammation. Lower Lysholm scores indicate greater knee-related symptoms. No differences in SF-12 or VAS pain scores were observed. The Lysholm score is a knee-specific metric of symptoms and functional disability (28). It is a weighted score, with pain and instability-related symptoms having the most weight (25 points each of a total 100). In contrast, the VAS scale only reflects knee pain, and the SF-12 health survey is not specific for knee-related issues. The unique association of inflammation with Lysholm scores and not VAS pain suggests that symptoms other than pain (e.g., instability, swelling) captured by the Lysholm scale might account for this difference. The weighting of the scale may also contribute to our observation. In the future, other knee-specific instruments such as the Knee Injury and OA Outcome Score (33), in which pain, other symptoms, and function can be independently evaluated, may be helpful in addressing this question. Whether inflammation is a cause or consequence of knee-specific symptoms in these patients (such as mechanical instability introduced by meniscal damage) needs to be evaluated.
We next looked at patient characteristics (age, BMI, degree of cartilage abnormality, and time elapsed between injury and surgery) in the stratified data. Age and BMI (34) are known risk factors for OA. In this cohort, older patients were more likely to demonstrate synovial inflammation, but BMI did not differ with inflammation. We anticipated that infiltration of cells would increase with time elapsed between injury and surgery, but this did not appear to be true. Patients with inflammation tended to have shorter time intervals between injury and surgery. A possible explanation is that increased inflammatory symptoms prompt earlier intervention; however, the present analysis did not address this issue. Multivariate analysis indicated that the association between inflammation and Lysholm scores is independent of age, BMI, and interval between injury and surgery.
Our results cannot be generalized to all patients with meniscal tears. We studied a population in which an identifiable injury precipitated symptoms, and in which injuries did not involve the vascular portion of the meniscus. Also, despite a clear history of trauma, most patients exhibited complex meniscal lesions. Although we excluded patients with radiographic OA, most patients demonstrated grade 1–4 Outerbridge cartilage lesions, suggesting that this population is enriched for patients with preradiographic disease. These observations show the presence of early joint degeneration in the majority of these patients (for review, see ref.35).
Synovial inflammation is associated with the progression of cartilage loss in patients with established OA (11, 12); thus, we determined whether inflammation was related to the degree of cartilage abnormality. There was a trend toward greater inflammation in patients with cartilage abnormalities, but our multivariate model demonstrated that the association between inflammation and Lysholm scores was independent of the degree of cartilage abnormality. Our finding of synovial inflammation in 1 of 7 patients with normal cartilage suggests that, in some cases of meniscal injury, synovitis may predate cartilage changes. This finding is consistent with those of an earlier study that demonstrated synovial immune complex deposits in patients with normal cartilage who were undergoing meniscal surgery (36). It is possible that synovitis contributes to alterations in structural and mechanical properties of meniscal tissues, resulting in susceptibility to meniscal injury and an increased risk of development or progression of OA.
To obtain insight into molecular mediators that contribute to synovial inflammation, we performed microarray analysis of synovial RNA. Four biopsy specimens from patients with inflammation (grade 1 or 2) and 4 from patients without (grade 0) were compared. Two hundred sixty genes were differentially expressed between these 2 patient groups (≥2-fold change). Pathway analysis, with a focus on genes involved in innate immune responses (31), revealed a set of chemokine and chemokine receptors among the most highly up-regulated transcripts in biopsy specimens with inflammation. Expression of these genes (Table 3) within synovium may promote recruitment of inflammatory cellular infiltrates, so we focused on this gene set for validation by real-time PCR.
We chose 5 genes for validation by real-time PCR: IL-8, CCL5, CCR7, CCL19, and CCL21. With the exception of IL-8, these belong to the “C-C” chemokine family, which influences the recruitment of monocytes, lymphocytes, and eosinophils. IL-8, a “C-X-C” chemokine, promotes neutrophil chemotaxis to sites of inflammation. Although first described as a T lymphocyte recruitment factor, CCL5 (or RANTES) has pleiotropic effects on multiple leukocyte subsets. CCR7 is the cognate receptor for both CCL19 and CCL21, which are involved in T lymphocyte and dendritic cell migration. Interaction between these chemokines and their receptor mediates homing to secondary lymphoid tissues and appropriate migration of cells within lymphoid follicles (for review, see ref.37). Our analysis revealed that IL-8, CCR7, and CCL5 transcripts were often undetectable in specimens without inflammation (Figure 2), and GEE analysis demonstrated that relative expression levels of CCL5 and CCL19 were associated with inflammation, consistent with our microarray results. Levels of CCR7 and CCL19 transcripts, which represent a ligand/receptor pair, were strongly associated with Lysholm scores (Figure 3).
There are limitations to analyses of large data sets, such as those obtained by microarray analysis, as they are prone to false-positive results and are not easily replicated (38). In the present study, our application of pathway analysis to this expression data set lends face validity to our findings, as genes are clustered functionally as well as statistically. Furthermore, chemokines identified with this high-throughput technique were validated using the more accurate method of quantitative PCR applied to a larger set of patients. Still, the biomarker potential of the chemokine signature needs to be validated in larger, prospective studies to determine whether these gene expression profiles have any diagnostic or prognostic biomarker potential. Patients enrolled in this study are being followed up for 2 years, and their clinical course assessed to determine whether the inflammatory response and/or chemokine expression is associated with short- and long-term outcomes after arthroscopy. Further limitations of our present study include the small sample size and the cross-sectional design. However, given the role of these chemokines in the recruitment of inflammatory cells, we speculate that they may contribute to the development of synovial inflammation in response to meniscal injury. Determining whether they directly affect the development of pain or progression of cartilage damage will require further investigation.
Identification of cellular and molecular mechanisms associated with synovial inflammation is of considerable interest, not only for the development of potential diagnostic or prognostic markers in early symptomatic OA and meniscal injury, but also for the development of therapeutic approaches to control clinical symptoms and potentially reduce the risk of joint degeneration in patients with knee injuries. Our study provides insight into mechanisms driving inflammatory infiltration and demonstrates an association between synovial inflammation and clinical symptoms in patients with meniscal injury, irrespective of the presence of underlying cartilage degeneration.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Scanzello had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Scanzello, McKeon, Richmond, Katz, Crow, Goldring.
Analysis and interpretation of data. Scanzello, McKeon, DiCarlo, Lee, Crow, Goldring.
The authors would like to acknowledge Fae Williams for clinical coordination of portions of the project, Kumar Bharat Rajan, PhD, for biostatistical support, Tibor Glant, MD, PhD, and Anna Laszlo for guidance in microarray data analysis, and David Hunter, MD, for input into the histologic score validation. In addition, we thank the Weill Cornell Medical College Microarray Core Facility for technical expertise in performing microarray data collection.